Correct Bank timestamp drift

[CriesofCarrots]

Problem

The timestamp returned by Bank::unix_timestamp() ends up in the clock sysvar and is used to assess time-based stake account lockouts. However it's based on a theoretical slots-per-second instead of reality, so it's quite inaccurate. This will pose a problem for lockups, since the accounts will not register as lockup-free on (or anytime near) the date the lockup is set to expire.

Block times are estimated using a validator timestamp oracle; this data provides an opportunity to align the bank timestamp more closely with real-world time.

Summary of Changes

• Add feature-gated EpochTimestamps sysvar that stores:
  1. The timestamp at the beginning of the epoch. Use it for updates to the clock sysvar, so that drift only happens from the beginning of the epoch. (Bank::unix_timestamp_from_epoch)
  2. A selection of slots in the upcoming epoch to sample.
• Add Bank method to get a bank's stake-weighted timestamp estimate using epoch vote accounts last_timestamp.
• Add Bank::update_epoch_timestamps:
  1. On Bank::new_from_parent, check the expected timestamp sample slots, and if one has passed, populate the sample with the stake-weighted timestamp.
  2. On epoch boundary, use populated sample timestamps to set a new epoch base timestamp in EpochTimestamps sysvar. Right now, all timestamp samples receive equal weight, regardless of when in the epoch they occurred. (TODO: Set sample slots for the upcoming epoch)

Right now, even with the feature activated this code is a no-op for the timestamp in the clock sysvar. Because the expected samples map is never populated, the timestamp at the epoch boundary defaults to the Bank::unix_timestamp_from_genesis().

TODO:

• Propose a method of selecting sampling slots
• Implement that method in Bank::update_epoch_timestamps

Fixes #9874
Toward #10093 (but requires some additional consideration of inflation timing)

[codecov]

Codecov Report

Merging #12681 into master will increase coverage by 0.0%.
The diff coverage is 97.9%.

@@           Coverage Diff            @@
##           master   #12681    +/-   ##
========================================
  Coverage    81.9%    82.0%            
========================================
  Files         359      361     +2     
  Lines       84158    84260   +102     
========================================
+ Hits        68994    69130   +136     
+ Misses      15164    15130    -34