[fix][client] Fix memory leak when message size exceeds max message size and batching is enabled

[lhotari]

Motivation

When enabling Netty leak detection with #23956 changes at paranoid level, BatchMessageTest.testSendOverSizeMessage shows a Netty ByteBuf memory leak. This is a bug in BatchMessageContainerImpl. It doesn't release the payload when the message size exceeds max message size.

Modifications

• Properly release the payload (or compressed payload) buffer when the message size exceeds max message size

Additional context

After #23956 changes are included, it is possible to catch such memory leaks in pull request CI builds.

Documentation

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

[codecov-commenter]

Codecov Report

All modified and coverable lines are covered by tests ✅

Project coverage is 74.23%. Comparing base (bbc6224) to head (9573768).
Report is 906 commits behind head on master.

Additional details and impacted files

@@             Coverage Diff              @@
##             master   #23967      +/-   ##
============================================
+ Coverage     73.57%   74.23%   +0.65%     
+ Complexity    32624    32260     -364     
============================================
  Files          1877     1853      -24     
  Lines        139502   143822    +4320     
  Branches      15299    16337    +1038     
============================================
+ Hits         102638   106762    +4124     
+ Misses        28908    28669     -239     
- Partials       7956     8391     +435     

Flag	Coverage Δ
inttests	26.72% <0.00%> (+2.13%)	⬆️
systests	23.22% <0.00%> (-1.10%)	⬇️
unittests	73.74% <100.00%> (+0.90%)	⬆️

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

Files with missing lines	Coverage Δ
.../pulsar/client/impl/BatchMessageContainerImpl.java	85.95% <100.00%> (+5.05%)	⬆️

... and 1040 files with indirect coverage changes