[To dev/1.3][Pipe] Fix pipe tablet memory self-lock during batching (#18266)

[Caideyipi]

Description

Backport of #18266 to dev/1.3.

A realtime Pipe can retain raw tablet events whose zero-sized PipeTabletMemoryBlock instances are accounted through forceResize. Before this fix, resize checked only remaining non-floating memory and could exhaust the pool, preventing the sink batch from allocating the memory needed to transfer and release those tablets.

This change applies the existing tablet/TsFile hard admission thresholds when a typed data-structure block grows through resize. Normal blocks, including sink batch buffers, keep the existing total-memory check.

The cherry-pick conflict was adapted to the 1.3 memory-manager implementation, which tracks usage through usedMemorySizeInBytes. The regression test uses the branch's configured non-floating memory size.

Tests

mvn test -pl iotdb-core/datanode -Dtest=PipeMemoryManagerResizeTest -DforkCount=0 -Dsurefire.forkNumber=0 -Ddevelocity.off=true

Result: 1 test run, 0 failures, 0 errors, 0 skipped. Checkstyle, Spotless, main compilation, and test compilation also passed.

---

This PR has:

• been self-reviewed.
• added comments explaining the intent of the code.
• added a unit test for the resized-tablet admission path.