Tracks last cleaned slot in accounts background service

[brooksprumo]

Problem

The main problem is #6295. As a means to that end, we need to ensure we don't clean past any enqueued snapshot request's slot. Currently, accounts background service tracks the last cleaned block height, which doesn't easily map to slot (when calling AccountsDb::clean_accounts(), which takes max_clean_root_inclusive, which is a Slot).

Summary of Changes

Track the last cleaned slot, instead of block height.

[codecov-commenter]

Codecov Report

Attention: Patch coverage is 61.53846% with 5 lines in your changes missing coverage. Please review.

Project coverage is 82.8%. Comparing base (39f3aa8) to head (7ec06e9).
Report is 1 commits behind head on master.

Additional details and impacted files

@@           Coverage Diff            @@
##           master    #6396    +/-   ##
========================================
  Coverage    82.8%    82.8%            
========================================
  Files         848      848            
  Lines      379048   379050     +2     
========================================
+ Hits       313859   314021   +162     
+ Misses      65189    65029   -160     

🚀 New features to boost your workflow:

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

[HaoranYi]

lgtm.

[brooksprumo]

Force-pushed to resolve merge conflicts. No code was changed.

[jeffwashington]

i'm not sure about skipped slots. we'll clean more often when we skip slots. Today we aren't skipping many slots on mnb. Just thinking about attacks, changes in behavior, etc.

[brooksprumo]

i'm not sure about skipped slots. we'll clean more often when we skip slots. Today we aren't skipping many slots on mnb. Just thinking about attacks, changes in behavior, etc.

Yep! This means we'll strictly clean more than before. Since mnb's skip rate is 0.2%, we'll only be changing the duration between calls to clean by like 1 slot. Note there's already a random amount applied to when we call 'clean' too. And really, when snapshots are enabled, we're cleaning each snapshot request, which have different (longer) runtimes than the no-snapshot-request paths.

Specifically, when snapshots are enabled (assuming the default 100 slot incremental snapshot interval), we never go through this else-if block, which is really the only thing that changed.

                        } else if bank.slot() - last_cleaned_slot
                            > (CLEAN_INTERVAL_SLOTS + thread_rng().gen_range(0..10))