[fix][ml] Don't estimate number of entries when ledgers are empty, return 1 instead

[lhotari]

Motivation

Following up on #24089 changes. The changes cause a regression which is visible in branch-3.0 where a test fails (org.apache.pulsar.compaction.CompactionTest#testDispatcherMaxReadSizeBytes). This test is different in newer branches. The regression is caused by the logic added in #24089 to use DEFAULT_ESTIMATED_ENTRY_SIZE + BOOKKEEPER_READ_OVERHEAD_PER_ENTRY as the default average entry size.
Another gap is that when all ledgers are empty, no average size would be used. That is also fixed in this PR

Modifications

• when the ledger is completely empty, don't estimate the number of entries, simply return 1 as the estimated entry count
• when all ledgers are empty beyond the read position, find the latest ledger with entries to calculate the average entry count to be used in the estimation
• add test coverage to cover these cases

Documentation

• doc
• doc-required
• doc-not-needed
• doc-complete

[codecov-commenter]

Codecov Report

Attention: Patch coverage is 84.61538% with 2 lines in your changes missing coverage. Please review.

Project coverage is 74.17%. Comparing base (bbc6224) to head (65f8a87).
Report is 990 commits behind head on master.

Files with missing lines	Patch %	Lines
...e/bookkeeper/mledger/impl/EntryCountEstimator.java	84.61%	1 Missing and 1 partial ⚠️

Additional details and impacted files

@@             Coverage Diff              @@
##             master   #24125      +/-   ##
============================================
+ Coverage     73.57%   74.17%   +0.59%     
+ Complexity    32624    32457     -167     
============================================
  Files          1877     1864      -13     
  Lines        139502   144452    +4950     
  Branches      15299    16472    +1173     
============================================
+ Hits         102638   107142    +4504     
+ Misses        28908    28844      -64     
- Partials       7956     8466     +510     

Flag	Coverage Δ
inttests	26.69% <46.15%> (+2.11%)	⬆️
systests	23.14% <46.15%> (-1.19%)	⬇️
unittests	73.68% <84.61%> (+0.83%)	⬆️

Flags with carried forward coverage won't be shown. Click here to find out more.

Files with missing lines	Coverage Δ
...e/bookkeeper/mledger/impl/EntryCountEstimator.java	93.44% <84.61%> (ø)

... and 1062 files with indirect coverage changes

🚀 New features to boost your workflow:

• ❄️ Test Analytics: Detect flaky tests, report on failures, and find test suite problems.
• 📦 JS Bundle Analysis: Save yourself from yourself by tracking and limiting bundle sizes in JS merges.