Complete the eval gate: completion regression, rerun, and multiple-comparison policies

## Current scope

PR #1027 completed the practical-significance work from this issue:

- [x] Exact one-sided sign test
- [x] 20% task-level net-win floor
- [x] Failed-slot judge retry and frozen successful judgments
- [x] Exact fail-closed result accounting

The remaining work is intentionally narrower:

- [ ] Add a trusted objective-completion regression hard gate when deterministic grader identity is available.
- [ ] Define the rerun policy so repeated runs cannot weaken the nominal significance guarantee.
- [ ] Define the multiple-comparison policy for suite-wide evaluation.

The completion gate remains deferred because the current infrastructure cannot identify a trusted deterministic completion grader. Scenario-level aggregation and any threshold changes must be back-tested before they become merge gates.

Related implementation: #1027.

_(Copilot, updating this issue on Abhitej's behalf.)_

Start by reviewing related PR #1027 and the current evaluation infrastructure. Determine how a trusted deterministic completion grader could be identified, then define rerun and suite-wide multiple-comparison policies; done requires these remaining policies and the completion regression gate to be specified and back-tested before becoming merge gates.

Written by the indexing model from the issue text.