Use compiled batches in replay advancement

[tradingexpert]

Summary

• add a C++ batch method that returns passive fills plus midpoint valuation marks
• make ReplayGateway.advance_to(...) use compiled batches when the engine supports them, with the existing scalar Python fallback preserved
• keep per-event valuation mark semantics while improving ordinary audited replay throughput
• rename the direct C++ per-event benchmark wording so it is clearly a compatibility baseline, not the fast path

Local Benchmarks

Direct engine:

• MatchingEngine scalar: 1,777,556 events/s
• CppMatchingEngine per-event: 1,857,812 events/s
• CppMatchingEngine batch: 15,964,170 events/s
• batch C++ speedup vs Python: 8.98x

Full audited replay:

• Replay + Python engine: 359,297 events/s
• Replay + default engine: 447,789 events/s
• default engine speedup vs Python: 1.25x

Tests

• python -m pip install -e ".[dev]"
• python -m ruff check .
• python -m pytest
• mkdir -p build/cpp && c++ -std=c++17 -Wall -Wextra -pedantic -Icpp cpp/tests/matching_engine_core_test.cpp -o build/cpp/matching_engine_core_test && build/cpp/matching_engine_core_test

Coverage: 98.20% total.