Remove BankStart

[apfitzge]

Problem

• BankStart is currently returned in BufferedPacketDecision
  • this forces us to take a lock on the poh_recorder to make decision
  • we want to get leader-bank w/o any locking

Summary of Changes

• Remove BankStart
  • this causes the removal several metrics related to leader bank detection time

Fixes #

[codecov-commenter]

Codecov Report

❌ Patch coverage is 91.17647% with 6 lines in your changes missing coverage. Please review.
✅ Project coverage is 83.0%. Comparing base (5d1651c) to head (33d2684).
⚠️ Report is 14 commits behind head on master.

Additional details and impacted files

@@           Coverage Diff            @@
##           master    #7351    +/-   ##
========================================
  Coverage    83.0%    83.0%            
========================================
  Files         800      800            
  Lines      362381   362262   -119     
========================================
- Hits       300792   300703    -89     
+ Misses      61589    61559    -30     

🚀 New features to boost your workflow:

• ❄️ Test Analytics: Detect flaky tests, report on failures, and find test suite problems.

[apfitzge]

Loss of metrics is main possible "issue" here, but along with #7355 this will let scheduler not lock poh at all.
I have not looked at these metrics in probably a year, I'm not sure if anyone has.

[tao-stones]

Not too worried about lost metrics either 😄 Before going into details, want to check with you if "not using bank_creation_time" is an intended improvement, not side effect.

[tao-stones]

Note: iirc, BankStart (mainly it's bank_creation_time) was added to limit banking_stage from packing too many transactions due to poh wall-clock drifting.

[tao-stones]

With block CU limits, i don't think overpacking is still an issue, at least not an issue to be mitigated with wall clock. So I think in general this is a positive change. Just want to call it out.

[apfitzge]

Before going into details, want to check with you if "not using bank_creation_time" is an intended improvement, not side effect.

It is more of a side-effect. I'd have likely just left it in if there was an easy way to do so.
We want a lock-free way to get the working bank from PoH, but I can't easily get the bank with that time; at least not in a clean way. The obj we'd have to store insidd the ArcSwapOption would be a Arc<(Arc<Bank>, Arc<Instant>)> instead of just an Arc<Bank> which seemed needlessly complicated for keeping the time to me.

[tao-stones]

lgtm