Add unit tests for core/architecture/sendsemantics partitioners

[Yicong-Huang]

Task Summary

Backfill unit tests for amber/src/main/python/core/architecture/sendsemantics/. This is the partitioner family that routes tuples between operators (range / hash / round-robin / broadcast / single / one-to-one) and is currently the largest under-tested module in the Python codebase per the latest Codecov report.

Codecov snapshot (commit f59e9e5, 2026-05-02):

file	cov	hits/lines
range_based_shuffle_partitioner.py	40.8%	20/49
broad_cast_partitioner.py	45.5%	20/44
round_robin_partitioner.py	46.1%	18/39
hash_based_shuffle_partitioner.py	47.5%	19/40
single_blocking_io.py (in core/models)	66.7%	36/54
module total	56.6%	133/235

102 lines unexecuted across 6 files. Goal: bring sendsemantics/ to ≥90% line coverage and ensure every partitioning branch is exercised.

Test ideas:

• Empty input batch routing for each partitioner
• Single-partition / single-receiver edge case
• Range partitioner: boundary keys, out-of-range keys, descending ranges
• Hash partitioner: hash collisions, null/None keys, deterministic routing across runs
• Round-robin: state continuity across multiple add() calls, restart behavior
• Broadcast: ensure tuple is emitted to all receivers, not duplicated to same receiver
• Backpressure / flush semantics if applicable to the partitioner contract
• Schema mismatch / type coercion errors

Out of scope: integration tests with the engine; end-to-end shuffle correctness tests live in amber/'s sbt test suite. This task is unit-test only, runs under pytest in the existing python CI job.

Task Type

• Testing / QA

Source: Codecov report at https://app.codecov.io/gh/apache/texera (commit f59e9e5e686fc34cbdfb8f4976706552cac01347).