ams-calibration: bounded, explicitly-approved self-adjustment of the min-rank threshold (mirrors #8121 discipline)

[JSONbored]

Problem

The end state of #8172: once the corpus, the advisory backtest, the track record, and the visibility surface all exist, the min-rank skip threshold can — with explicit operator approval — be allowed to self-adjust inside hard bounds, exactly as ORB's satisfaction floor does under #8121.

Requirements

⚠️ Required pattern. The full #8121 discipline, no shortcuts: double gating (an explicit opt-in config flag AND a per-apply approval), loosening-only movement inside a hard-bounded range declared next to the shipped constant, held-out split must independently clear, every apply writes a typed ledger event with the complete evidence, and a one-command revert.

• Config flag in .loopover-ams.yml (default OFF) mirroring how DEFAULT_AMS_POLICY_SPEC overrides already parse.
• Apply path validates bounds at the single consumption point (defaulted parameter, never a mutable global).
• Revert command restores the shipped default and records the reversion event.

Deliverables

• Double-gated apply + revert with 100%-branch coverage
• Ledger evidence events + doctor/status reflection
• Docs note in the miner's calibration docs section

Links & Resources

#8172 (epic), #8121 (the discipline), #8158/#8161 (ORB's cron + visibility analogs), packages/loopover-engine/src/ams-policy-spec.ts

Boundaries

Nothing autonomous without the explicit opt-in; one knob only; local node only.

maintainer-only — epic decomposition of #8172.

[linear-code]

JSO-5685