* Decouple `relays/client-substrate` from `headers_relay`
* Remove `blocks_in_state` from `SyncLoopMetrics`
    This metric was only relevant for PoA <> Substrate bridge.
* Move `sync_loop_metrics.rs` to `relays/finality`
* Remove unused `SyncLoopMetrics::update()`
* Hook up SyncLoopMetrics to finality_loop
* Delete now unused `relays/headers`

Signed-off-by: acatangiu <[email protected]>

* Unify error enums in substrate and ethereum clients with `thiserror`

Related to https://github.com/paritytech/parity-bridges-common/issues/857

* Add license pre-amble

* rustfmt

* Fix spelling

* Alter gitlab.

* Use substrate's rustfmt.toml

* cargo +nightly fmt --all

* Fix spellcheck.

* cargo +nightly fmt --all

* format.

* Fix spellcheck and fmt

* fmt?

* Fix spellcheck

Co-authored-by: Tomasz Drwięga <[email protected]>

* I'm just trying to do my job

* spelling

* read justifications from stream using channel + task that fills that channel

* Arc<Mutex> -> Mutex

* shared conversion rate metric value

* clippy

* CI: do not trust this CI 1

* CI: don't want to trigger unneded statuses

* CI: debug 1

* CI: new CI

* CI: temp allow failure

* CI: exclude paths

* doc: check if docs won't trigger pipelines

* doc: check hybrid changes pipeline

* doc: do not push excluded files together with others

* CI: undebug fmt

* CI: better deny

* CI: fix deny and spellcheck

* CI: global backtrace

* CI: deny config

* CI: publishing

* Dockerfile: metadata fix [skip ci]

* CI: revert me

* CI: debug bash

* CI: mv ci.Dockerfile; fix buildah bug

* CI: fix artifact name

* Dockerfile: fix context

* CI: separate deny check licenses

* CI: when to run

* CI: unneded stuff in these Dockerfiles

* CI: merged test-refs and build-refs

* CI: test-build optimizations

* CI: changes, web, scheduled pipelines now work as intended

* CI: use tested production CI image

* CI: substitute GHA

* Fix clippy.

* Moar clippy fixes.
...

* always run on-demand headers relay in complex relay

* fix compilation

* do not spawn additional task for on-demand relays

* compilation

* prefix in relay loops logs

* fmt

* complex headers+messages relay

* post-merge fix

* fix + test issue with on-demand not starting

* administrative metrics

* fmt

* fix compilation

* fix compilation again

* and another one

* remove GenericLoopMetrics

* chttp -> isahc

* remove redundant marker

* not about price metrics

* fmt

* Stop passing raw encoded justifications to pallet API

By having the API accept a struct-ified justification we are able to
better utilize the justifications fields for weight calculations.

* Update relayer code to use decoded justifications

* Add justification to `expect()` statement

* Fix some imports

* Make justification wrapper contain decoded justification

* Rename some fields

* Get rid of warnings

* Appease Clippy

* Only decode justification once at init time

* Remove unnecessary method

* Remove justification wrapper

This became kinda unnecessary since we could implement the FinalityProof
trait on GrandpaJustification directly.

* slightly changed relay loop initialization

* git mv

* clippy

* more clippy

* loop_run -> run_loop

* review and clippy

* clippy

* Remove chains.

* Move relay clients.

* Flatten generic.

* Fix fmt.