BankingStage: Gossip votes held

[apfitzge]

Problem

Migrating issue from solana-labs#34973

• Gossip votes are held in ClusterInfoVoteListener until we become leader.
• This seems unneccessary now since both voting threads insert into a shared data-structure which only keeps the latest vote.
• Holding the votes can lead to misleading metrics since tpu votes are a sustained throughput, whereas gossip votes come in nearly all at once

Proposed Solution

• ClusterInfoVoteListener just verifies the votes and sends to BankingStage for it to handle

[apfitzge]

@carllin @AshwinSekar - I migrated this issue. My understanding from previous issue is it is not necessary for us to continue doing this holding.

@carllin if that is not the case, please let us know.

[AshwinSekar]

Confirmed with carl in April that votes do not need to be held in ClusterInfoVoteListener with the caveat that banking stage only drains the votes for slots on the working_bank's fork and leaves the remaining untouched.

The benefit of holding votes in ClusterInfoVoteListener is that if the leader decides to build it's first block off of fork1 and it's second block off of fork2, gossip votes for fork2 will still make it into the second block, even if they are filtered out for the first block.

Essentially we want to move this check:

agave/core/src/verified_vote_packets.rs

Lines 89 to 90 in a2eaf6d

	// Filter out votes on the wrong fork (or too old to be)
	// on this fork

Into here:

agave/core/src/banking_stage/latest_unprocessed_votes.rs

Lines 322 to 323 in a2eaf6d

	/// Drains all votes yet to be processed sorted by a weighted random ordering by stake
	pub fn drain_unprocessed(&self, bank: Arc<Bank>) -> Vec<Arc<ImmutableDeserializedPacket>> {