Deduplicate blocksync

[kaimast]

Motivation

There are two different code paths that perform block synchronization: one for validators, and one for clients/provers.
In the case of validators, there exist two different instances of BlockSync.

This PR aims to address this by moving code that specific to validators or clients. It also improves documentation of the block synchronization logic.
Finally, the PR serves as the basis for upcoming changes to block synchronization that will harden security and improve performance of the protocol.

Overview

The following is an outline of the structs affected by this PR. It also highlights which structs received significant changes.

Changes

There have been a multiple updates to this PR, and I would prefer if it gets squashes before a merge to not clutter the commit history too much.

Below is a list of the first set of commits that implemented the vast majority of the PR.

• 509c266 changes the validator code to only use one instance of BlockSync.
• 258ffcb and 173e8e4 perform minor clean-ups and improve code documentation/comments.
• 4adaf15 removes duplicated code from testing, because we need to pass BlockSync every time due to the very first commit.
• Finally, 0c75996 significantly refactors BlockSync and moves all client specific code to the client.

Test Plan

Existing tests should suffice, as this PR does not change any functionality.

[howardwu]

^You can add this to the main PR description for improved visibility.

Follow up questions for main PR description:

• What memory or time performance improvements to syncing (if any) are improved in this PR?
• What unit or integration tests were introduced to validate the changes made to syncing in this PR?

[raychu86]

@kpandl Could you run the sync tests on this PR to check if there are any improvements/regressions to syncing after the change?

[kaimast]

I updated the PR description to better reflect the current state of this PR and added a simplified diagram. Hope we can get it merged this week. @vicsn

[raychu86]

Great cleanup and unification of disjoint usages! Code overall looks logically correct.

Would wait to merge until proper sync stress tests have been run and reported. @kpandl would appreciate any updates once you've had a change to run the tests

[kpandl]

I've ran 2 clients syncing on mainnet.

Yellow is staging, and green is this feature branch. So syncing generally works. For this feature branch, syncing was ~3% slower, so not much. Would need more runs to have significance, but since this PR mostly reorganizes, might be noise. LGTM.

Note that I haven't run the classic stress-test, since we are currently having issues with the automated test infra.