fix(validation): restore LEAN case-study benchmark

[stefan-jansen]

Summary

• restore the LEAN case-study parity harness
• commit the required compressed/split LEAN reference fixtures so CI runs the benchmark instead of skipping it
• hard-fail when reference artifacts are missing
• clear direct cost fields in zero-cost benchmark configs and report LEAN decoded fills as trade counts

Validation

• pre-commit run --all-files
• uv run pytest tests/benchmark/test_case_study_lean.py -q
• uv run pytest tests/ -q
• clean worktree: uv run pytest tests/benchmark/test_case_study_lean.py -q

[Copilot]

Pull request overview

Restores and hardens the LEAN validation/benchmarking surface so cross-engine parity comparisons remain consistent, and adds a LEAN case-study regression test + harness code under ml4t.backtest._validation.

Changes:

• Add helpers to fully disable commission/slippage fields in validation/benchmark_suite.py to prevent cost-model auto-activation from preset leftovers.
• Adjust LEAN benchmark reporting to treat decoded fills as “trade count” for parity with other engines.
• Add a LEAN case-study parity harness module and a regression test that compares fill multisets + terminal portfolio value.

Reviewed changes

Copilot reviewed 3 out of 3 changed files in this pull request and generated 3 comments.

File	Description
validation/benchmark_suite.py	Centralizes “zero-cost mode” resets and adjusts LEAN trade-count reporting for cross-engine parity.
tests/benchmark/test_case_study_lean.py	Adds a regression test intended to validate ml4t’s LEAN reconstruction against committed LEAN artifacts.
src/ml4t/backtest/_validation/case_study_lean.py	Implements the case-study harness: workspace parsing, LEAN zip decoding, ml4t replay, and parity comparison utilities.

---

💡 Add Copilot custom instructions for smarter, more guided reviews. Learn how to get started.