feat: Support watermark status propagation

[lgbo-ustc]

Summary

• add native WatermarkStatus stream element propagation
• track combined idle/active watermark status across inputs
• expose StatefulTask/StatefulOperator APIs for watermark status notifications
• add CombinedWatermarkStatus unit coverage for active/idle transitions

releated PRs

• Support watermark status propagation velox4j#49
• [GLUTEN-12397][FLINK] Support WatermarkStatus propagation apache/gluten#12398

Validation

• git diff --check
• cmake --build build --target velox_stateful_combined_watermark_status_test -j 8
• ./build/velox/experimental/stateful/tests/velox_stateful_combined_watermark_status_test
• velox4j C++/Java compile passed after updating velox4j to this Velox change

Note

• standalone velox_stateful_exec build was previously blocked because the environment was missing xsimd/xsimd.hpp through unrelated stateful dependencies; the new CombinedWatermarkStatus test target avoids those unrelated dependencies and passed.

[KevinyhZou]

int -> int32_t

[KevinyhZou]

ditto

[lgbo-ustc]

Addressed review feedback: changed int to int32_t for the watermark index parameters across the call chain (CombinedWatermarkStatus, StatefulOperator, StatefulTask) in commit bb19102.