The invocation is a bit more verbose than I'd prefer, but it's also
more explicit about what types need to be public. I'll take it as a win.

Unfortunately, it doesn't work right:

```
error[E0446]: private type `CandidateBackingJob` in public interface
   --> node/core/backing/src/lib.rs:775:1
    |
86  | struct CandidateBackingJob {
    | - `CandidateBackingJob` declared as private
...
775 | delegated_subsystem!(CandidateBackingJob as CandidateBackingSubsystem);
    | ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ can't leak private type
```

I'm not sure precisely what's going wrong, here; I suspect the problem is
the use of `$job as JobTrait>::RunArgs` and `::ToJob`; the failure would be
that it's not reifying the types to verify that the actual types are public,
but instead referring to them via `CandidateBackingJob`, which is in fact private;
that privacy is the point.

Going to see if I can generic my way out of this, but we may be headed for a
quick revert here.

* chain-api subsystem skeleton

* chain-api subsystem: draft impl

* chain-api subsystem: mock testclient

* chain-api subsystem: impl HeaderBacked for TestClient

* chain-api subsystem: impl basic tests

* chain-api subsystem: tiny guide

* chain-api subsystem: rename ChainApiRequestMessage to ChainApiMessage

* chain-api subsystem: add the page to the ToC

* chain-api subsystem: proper error type

* chain-api subsystem: impl ancestors request

* chain-api subsystem: tests for ancestors request

* guide: fix ancestor return type

* runtime-api subsystem: remove unused dep

* fix fmt

* fix outdated comment

* chain-api subsystem: s/format/to_string

* lower-case subsystem names

* chain-api subsystem: resolve Finalized todo

* chain-api subsystem: remove TODO

* extract request errors into a module

* remove caching TODO

* fix imports

Subsystems are encouraged to either typedef themselves as appropriate
`JobManager` instances for their job type, or wrap a `JobManager`
instance and delegate the `Subsystem` impl. In both cases, we want
to use a sensible, non-repeated subsystem name for appropriate
logging and debugging.

This PR adds a heuristic: if the job name ends in the literal
"Job", then that gets stripped. Otherwise, the job name is used.
This improves on the previous situation, in which subsystems
typedef'd to or wrapping `JobManager` all got the same constant (!)
name.

* polkadot-subsystem: update runtime API message types

* update all networking subsystems to use fallible runtime APIs

* fix bitfield-signing and make it use new runtime APIs

* port candidate-backing to handle runtime API errors and new types

* remove old runtime API messages

* remove unused imports

* fix grumbles

* fix backing tests

* Initial commit

* WIP

* Make atomic transactions

* Remove pruning code

* Fix build and add a Nop to bridge

* Fixes from review

* Move config struct around for clarity

* Rename constructor and warn on missing docs

* Fix a test and rename a message

* Fix some more reviews

* Obviously failed to rebase cleanly

* add ActiveLeavesUpdate, remove StartWork, StopWork

* replace StartWork, StopWork in subsystem crate tests

* mechanically update OverseerSignal in other modules

* convert overseer to take advantage of new multi-hash update abilities

Note: this does not yet convert the tests; some of the tests now freeze:

test tests::overseer_start_stop_works ... test tests::overseer_start_stop_works has been running for over 60 seconds
test tests::overseer_finalize_works ... test tests::overseer_finalize_works has been running for over 60 seconds

* fix broken overseer tests

* manually impl PartialEq for ActiveLeavesUpdate, rm trait Equivalent

This cleans up the code a bit and makes it easier in the future to
do the right thing when comparing ALUs.

* use target in all network bridge logging

* reduce spamming of  and

* Companion for #6726

* Spaces

* 'Update substrate'

Co-authored-by: parity-processbot <>

* get conclude signal working properly; don't allocate a vector

* wip: add test suite / example / explanation for using utility subsystem

Unfortunately, the test fails right now for reasons which seem
very odd. Just have to keep poking at it.

* explicitly import everything

* fix subsystem-util test

The root problem here was two-fold:

- there was a circular dependency from subsystem -> test-helpers/subsystem ->
  subsystem
- cfg(test) doesn't propagate between crates

The solution: move the subsystem test helpers into a sub-module
within subsystem. Publicly export them from the previous location
so no other code breaks.

Doing this has an additional benefit: it ensures that no production
code can ever accidentally use the subsystem helpers, as they are compile-
gated on cfg(test).

* fully commit to moving test helpers into a subsystem module

* add some more tests

* get rid of log tests in favor of real error forwarding

It's not obvious whether we'll ever really want to chase down
these errors outside a testing context, but having the capability
won't hurt.

* fix issue which caused test to hang on osx

* only require that job errors are PartialEq when testing

also fix polkadot-node-core-backing tests

* get rid of any notion of partialeq

* rethink testing

Combine tests of starting and stopping job: leaving a test executor
with a job running was pretty clearly the cause of the sometimes-hang.

Also, add a timeout so tests _can't_ hang anymore; they just fail
after a while.

* rename fwd_errors -> forward_errors

* warn on error propagation failure

* fix unused import leftover from merge

* derive eq for subsystemerror

* Use SpawnNamed instead of Spawn in Overseer

* reexport SpawnNamed and fix doc tests

* Fix deps

* Add subsystem-util crate.

Start by moving the JobCanceler here.

* copy utility functions for requesting runtime data; generalize

* convert subsystem-util from crate to module in subsystem

The point of making a sub-crate is to ensure that only the necessary
parts of a program get compiled; if a dependent package needed only
subsystem-util, and not subsystem, then subsystem wouldn't need to
be compiled.

However, that will never happen: subsystem-util depends on
subsystem::messages, so subsystem will always be compiled.

Therefore, it makes more sense to add it as a module in the existing
crate than as a new and distinct crate.

* make runtime request sender type generic

* candidate backing subsystem uses util for api requests

* add struct Validator representing the local validator

This struct can be constructed when the local node is a validator;
the constructor fails otherwise. It stores a bit of local data, and
provides some utility methods.

* add alternate constructor for better efficiency

* refactor candidate backing to use utility methods

* fix test breakage caused by reordering tests

* restore test which accidentally got deleted during merge

* start extracting jobs management into helper traits + structs

* use util::{JobHandle, Jobs} in CandidateBackingSubsystem

* implement generic job-manager subsystem impl

This means that the work of implementing a subsystem boils down
to implementing the job, and then writing an appropriate
type definition, i.e.

pub type CandidateBackingSubsystem<Spawner, Context> =
	util::JobManager<Spawner, Context, CandidateBackingJob>;

* add hash-extraction helper to messages

* fix errors caused by improper rebase

* doc improvement

* simplify conversion from overseer communication to job message

* document fn hash for all messages

* rename fn hash() -> fn relay_parent

* gracefully shut down running futures on Conclude

* ensure we're validating with the proper validator index

* rename: handle_unhashed_msg -> handle_orphan_msg

* impl Stream for Jobs<Spawner, Job>

This turns out to be relatively complicated and requires some
unsafe code, so we'll want either detailed review, or to choose
to revert this commit.

* add missing documentation for public items

* use pin-project to eliminate unsafe code from this codebase

* rename SenderMessage -> FromJob

* reenvision the subsystem requests as an extension trait

This works within `util.rs`, but fails in `core/backing/src/lib.rs`,
because we don't actually create the struct soon enough. Continuing
down this path would imply substantial rewriting.

* Revert "reenvision the subsystem requests as an extension trait"

This reverts commit a5639e36.

The fact is, the new API is more complicated to no real benefit.

* apply suggested futuresunordered join_all impl

* CandidateValidationMessage variants have no top-level relay parents

* rename handle_orphan_msg -> handle_unanchored_msg

* make most node-core-backing types private

Now the only public types exposed in that module are
CandidateBackingSubsystem and ToJob. While ideally we could reduce
the public interface to only the former type, that doesn't work
because ToJob appears in the public interface of CandidateBackingSubsystem.

This also involves changing the definition of CandidateBackingSubsystem;
it is no longer a typedef, but a struct wrapping the job manager.