revert(ml): keep head auto-apply precision at 0.97 (operator: general tuning was fine)

Milestone 139 raised head_auto_apply_precision 0.97→0.98; operator confirmed the
general-tag confidence was already well tuned, so revert that. The support floor
(min_positives 30→50) and CCIP match confidence (0.92→0.95) stay. Migration 0081
(not yet deployed) edited to drop the precision bump.

Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01CDgx8bQS5YrGRK76v8HUnM

backend/app/models/ml_settings.py

@@ -51,10 +51,7 @@ class MLSettings(Base):
         Integer, nullable=False, default=8
     )
     head_auto_apply_precision: Mapped[float] = mapped_column(
-        # Stricter graduation bar (was 0.97) to cut auto-apply misfires
-        # (operator-asked 2026-07-06): a higher precision target → fewer heads
-        # graduate and those that do get a higher per-head auto_apply_threshold.
-        Float, nullable=False, default=0.98
+        Float, nullable=False, default=0.97
     )
     # Earned auto-apply (#114). A graduated head fires (tags images without a
     # human) when this master switch is on AND the head has at least