Issue 511 — Lock DepMap reporting boundaries and refresh README

[rsolerortuno]

Summary

Implements Issue 511: DepMap reporting regression, isolation, and README refresh.

• rebuilds README.md around the current v0.5.0 research-preview architecture;
• restores benchmark, sensitivity, and external-validation boundaries;
• locks legacy Markdown and HTML compatibility;
• validates optional evidence routing and deterministic composition;
• proves dataframe, score, rank, and role non-mutation;
• proves the productive pipeline remains unwired from DepMap evidence;
• locks filesystem, environment, network, snapshot, and matrix isolation.

Scientific boundaries

• The productive baseline remains the original 300-gene ranking.
• The 56-target benchmark is internally curated.
• The 331-identity discovery universe remains distinct from the productive baseline.
• The 18,531-gene DepMap universe is background only.
• DepMap evidence is optional and post-ranking.
• Baseline scores and ranks remain unchanged.
• Production activation remains disabled.
• Human review remains mandatory.
• DepMap dependency is not clinical anti-PD-1 response evidence.
• Lack of tumor-cell dependency does not invalidate an immune target.
• Real DepMap Public 26Q1 snapshot publication remains pending Issue 512.

Validation

• New regression, isolation, and README tests passed.
• Historical README metric and v0.3.0 boundary tests passed.
• Full repository suite passed.
• AGY adversarial audit completed.
• Claude review approved.
• git diff --check passed.

No production code was modified.