implement custom proposer (#1320)

* network bridge skeleton

* move some primitives around and add debug impls

* protocol registration glue & abstract network interface

* add send_msgs to subsystemctx

* select logic

* transform different events into actions and handle

* implement remaining network bridge state machine

* start test skeleton

* make network methods asynchronous

* extract subsystem out to subsystem crate

* port over overseer to subsystem context trait

* fix minimal example

* fix overseer doc test

* update network-bridge crate

* write a subsystem test-helpers crate

* write a network test helper for network-bridge

* set up (broken) view test

* Revamp network to be more async-friendly and not require Sync

* fix spacing

* fix test compilation

* insert side-channel for actions

* Add some more message types to AllMessages

* introduce a test harness

* impl ProvideInherent for InclusionInherent

* reduce import churn; correct expect message

* move inclusion inherent identifier into primitives

It's not clear precisely why this is desired, but it's a pattern
I've seen in several places, so I'm going this to be on the
safe side. Worst case, we can revert this commit pretty easily.

* bump kusama spec_version to placate CI

* copy sc_basic_authorship::{ProposerFactory, Proposer}

We have from the problem description:

> This Proposer will require an OverseerHandle to make requests via.

That's next on the plate.

* use polkadot custom proposer instead of basic-authorship one

* add some tests

* ensure service compiles and passes tests

* fix typo

* fix service-new compilation

* Subsystem test helpers send messages synchronously

* remove smelly action inspector

* remove superfluous let binding

* fix warnings

* add license header

* empty commit; maybe github will notice the one with changes

* Update node/network/bridge/src/lib.rs

Co-authored-by: Peter Goodspeed-Niklaus <coriolinus@users.noreply.github.com>

* add sanity check to only include valid inherents

* stub: encapsulate block production mechanics instead of copying them

The goal is to end up with something like what's in
validation::block_production::*, which encapsulates
basic block production mechanics. This is a better idea than
just straight-up copying those mechanics.

* partial implementation of propose fn

Doesn't actually compile yet; need to bring in some other
commits to ensure ProvisionerMessage is a thing, and also
figure out how to get the block hash given the current
context.

* fix compilation

* clear a few more compile errors

* finish fn propose

* broken: add timeout to proposal

* add timeout to proposal

* guide: provisioner is responsible for selecting parachain candidates

* implement ProvisionerMessage::RequestInherentData & update fn propose

* impl CreateProposer::init; clean up

* impl std::error::Error for Error

* document error-handling rationale

* cause polkadot-service-new to compile correctly

* Move potentially-blocking call from fn init -> fn propose

This means that we can wrap the delayed call into the same
timeout check used elsewhere.

* document struct Proposer

* extract provisioner data fetch

This satisfies two requirements:

- only applies the timeout to actually fetching the provisioner data,
  not to constructing the block after
- simplifies the problem of injecting default data if we could not
  get the real provisioner data in time.

Co-authored-by: Robert Habermeier <rphmeier@gmail.com>
Co-authored-by: Gavin Wood <gavin@parity.io>