fix: correct auto-grade adjustment bounds in docstrings

[AGRO-CODEX]

What

Corrects two docstrings in helpers/grade.py that claimed auto-grade adjustments are "capped at +-8% on any axis." The actual bounds differ per axis.

Why

The real clamps in auto_grade_for_clip are:

contrast_adj = max(0.94, min(1.08, contrast_adj))  # [-6%, +8%]
gamma_adj    = max(0.94, min(1.10, gamma_adj))      # [-6%, +10%]
sat_adj      = max(0.94, min(1.06, sat_adj))        # [-6%, +6%]

Gamma can go up to +10% for underexposed clips, not +8%. The module-level docstring and the auto_grade_for_clip docstring both stated "+-8%" which is misleading — a reader checking the docs before tuning the caps would see the wrong range.

Changes

• Module docstring: replaced "All adjustments capped at +-8% on any axis" with per-axis bounds.
• auto_grade_for_clip docstring: same correction.

No logic changes.

---

Summary by cubic

Update module and auto_grade_for_clip docstrings to state the real per-axis bounds for auto-grade adjustments: contrast ±8%, gamma +10%/-6%, saturation ±6%. Fixes the misleading “±8% on any axis” wording; no behavior changes.

^{Written for commit e1375ad. Summary will update on new commits.}

[cubic-dev-ai]

No issues found across 1 file

_{Re-trigger cubic}