review: unit-test pure helpers for the review/trail commands

These deterministic helpers added on this branch sat at 0% coverage. Add focused unit tests (no I/O, no interactive forms, no network) covering:

• review: finalJudgeDisplayName, exampleAgentList, nonLaunchableEligibleNames, launchableInstalledAgentNames, slotLabel, defaultAgentPick, filterOutBuiltinCollisions, dedupeStrings
• cli (trail): severityDisplay, trailReviewTargetDisplay, defaultTrailReviewStatusReason, parseOptionalTrailSelector, truncateForLog

All 13 helpers now at 100%. The interactive pickers, RunE handlers, and git/SSE side-effecting functions remain integration-test territory.