Rework p2p req resp protocol

[anacrolix]

Primarily addresses #11779.

• Adds the ability to configure the SyncClient peer rate limit and ReqRespServer global rate limit via the SetupP2P interface.
• Adds a test that has a bunch of syncers connect to a sequencer after a large number of blocks have been lost to gossip.
• Adds PayloadSource to L2PayloadIn handlers, so it can be differentiated between gossip and alt sync.
• Changes GossipIn to L2PayloadIn in a few places where it's not just gossip.
• Removes non-functional/unused active range request tracking. There's just one gap at a time.

Reworks p2p alt sync to:

• Prioritize the next trusted block and blocks that can be promoted soonest.
• Adds request concurrency so a single peer can't block progression.
• Retry failed requests to other peers.
• Correctly wait if an error is received from a server.
• Avoid overflowing quarantine and trusted caches.
• Always promote the very next trusted block.
• Handle range request updates without throwing away or forgetting about existing requests.

TODO/open questions/future PRs (not blocking unless someone flags them):

• Should I add a metric in OpNode where it can differentiate the L2 unsafe payload source? Do we have a precedent for handling error and non error cases (separate metrics?).
• Limit the number of blocks to quarantine, and evict blocks furthest from the promote head.
• Add a rate limit error code to the req/resp protocol so clients don't penalize servers for rate limiting them.
• Trigger range request updates when the unsafe head moves. I put the code in place for this but it doesn't seem to trigger.
• The active request block list could be a slice instead of a map.
• Possibly handle abandoning the gap entirely when we get an empty range request.

[anacrolix]

The contracts-bedrock-tests failure is unrelated I think...

This is ready for review.

[anacrolix]

Thanks @tynes, addressed your nits, and answered the TODO question so it's not lingering forever as an unknown.

[tynes]

How important is it to use these packages? We need to be really sure when adding new packages to be certain there aren't backdoors

[anacrolix]

I definitely need chansync or I'd just end up duplicating all that again.

Some of the generic stuff is exceptionally useful for avoiding a lot of verbosity in Go.

The rest I was using for debugging.

[semgrep-app]

Please create a GitHub ticket for this TODO.

_{Ignore this finding from todos_require_linear.}

[semgrep-app]

Semgrep found 1 nil-check-after-call finding:

• op-node/p2p/node.go
  • L99-191 - Triage

Potential p2pmetrics nil dereference when NewBandwidthCounter is called

_{Ignore this finding from nil-check-after-call.}