Fix topic reader flow-control stall on partition session stop

[Eistern]

When ReadPartitionSession.stop() was called (e.g. on graceful partition eviction) any batches still sitting in readingQueue had their readFuture abandoned. Those futures are part of CompletableFuture.allOf in ReadSession#onReadResponse; if they never complete, sendReadRequest() is never invoked for that ReadResponse, the SDK never replenishes the server-side memory budget, and the whole read session stops requesting data.

This restores the drain that used to live in v2.3.33 PartitionSessionImpl .shutdown() and was lost during the refactor that introduced MessageCommitter (b862816). It also closes a small race where stop() runs between the entry check and readingQueue.offer() in addBatches().

Adds a regression test that reproduces the stall on master and verifies the fix.

[codecov-commenter]

Codecov Report

❌ Patch coverage is 50.00000% with 3 lines in your changes missing coverage. Please review.
✅ Project coverage is 71.54%. Comparing base (fdecb35) to head (6204d74).

Files with missing lines	Patch %	Lines
...tech/ydb/topic/read/impl/ReadPartitionSession.java	50.00%	2 Missing and 1 partial ⚠️

Additional details and impacted files

@@             Coverage Diff              @@
##             master     #657      +/-   ##
============================================
+ Coverage     71.38%   71.54%   +0.15%     
- Complexity     3375     3379       +4     
============================================
  Files           379      379              
  Lines         15933    15938       +5     
  Branches       1670     1672       +2     
============================================
+ Hits          11374    11403      +29     
+ Misses         3909     3890      -19     
+ Partials        650      645       -5     

☔ View full report in Codecov by Sentry.
📢 Have feedback on the report? Share it here.

🚀 New features to boost your workflow:

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