fix: replace Forager quality-gate hardcoded literal with named constant

[keitaj]

Summary

• Replace the bare 50.0 literal in Forager's quality-gate with a reference to CoinHealthTracker._NO_HISTORY_NEUTRAL_SCORE, the named midpoint the tracker already uses for unknown coins.
• The two values were always meant to track each other — pin the dependency explicitly so one location is the source of truth for "neutral activity".
• Add boundary tests that pin the strict-inequality semantic of the gate so a future refactor of > to >= is caught by CI.

Addresses the should-fix from PR #146 review (#146 (review)).

Changes

• strategies/market_making_strategy.py L1208-1219: swap the literal 50.0 for CoinHealthTracker._NO_HISTORY_NEUTRAL_SCORE. Update the surrounding comment to document the linkage.
• tests/test_forager.py: add test_quality_gate_boundary_at_neutral_score and test_quality_gate_just_above_neutral_score_blocks_trigger to pin the strict > boundary.

Test plan

• flake8 . --max-line-length=120 passes
• pytest tests/ -q passes (1006 tests, 0 failures — 2 new under test_forager.py)
• No behaviour change with forager_enabled=false (the gate is unreachable)
• No behaviour change with default forager_enabled=true either — the value emitted by the named constant equals the prior literal exactly

🤖 Generated with Claude Code

[keitaj]

Review: Replace Forager quality-gate hardcoded literal with named constant

Surgical follow-up to the should-fix from PR #146 review. The literal 50.0 is replaced with CoinHealthTracker._NO_HISTORY_NEUTRAL_SCORE, the comment now documents why those two values are linked, and two boundary tests pin the strict-inequality semantic so a future > → >= refactor is caught by CI. No runtime behaviour change.

---

1. Cross-module access to a leading-underscore attribute (nit)

strategies/market_making_strategy.py L1219:

CoinHealthTracker._NO_HISTORY_NEUTRAL_SCORE is referenced from outside the class. The leading underscore is the Python convention for "private", and now there are two consumers (strategy + the new boundary tests) treating it as public.

Suggestion (optional, future cleanup): promote the attribute to NO_HISTORY_NEUTRAL_SCORE (no leading underscore) when other Forager-internal constants accumulate, or expose it via a small accessor (@classmethod neutral_score()). Not needed for this PR — the underscore convention is informational and there's no real encapsulation barrier in Python — but worth noting for the next refactor pass.

---

Verdict: LGTM

No blockers, no should-fix. The change is minimal, deterministic, and adds regression-pinning tests for the boundary semantic. Ready to merge.

🤖 Generated with Claude Code