Outcome 5 (2/2): eval run --replicate K with statistical report

[DianaMeda]

Building block: stacked on #175

⚠️ Stacked on #175 (Outcome 5 1/2: replication statistics aggregator). Merge #175 first. Until then this PR's diff includes #175's commit; once #175 lands in main, this reduces to just the 7 files here.

Why

Completes Outcome 5: turns the pure aggregator (#175) into a single command that produces a statistically-grounded comparison artifact — lift as mean ± SD with a one-SD significance flag — exactly the GOAL.md Constraint that replication be "automatable as a single eval run --replicate K command."

What's here

• cli/commands/run.ts — --replicate <K> (default 1). K<1 errors; 2 ≤ K < 5 warns that K≥5 is the minimum for a credible signal. For each harness the A/B comparison runs K times; the K ComparisonResults are aggregated via aggregateReplications() into a ReplicationReport, persisted by runId, and printed. K=1 leaves the existing single-run output/behavior unchanged.
• storage — saveReplicationReport(runId, report) / getReplicationReport(runId) on the ResultStore interface and the JSON implementation (runs/<runId>/replication.json, retrievable after the process exits).
• output/replication-display.ts — formatReplicationReport(): per dimension, mean lift ± SD, applicable replications, t-statistic, and SIGNAL/none.

Notes

• Sequential execution (parallel/interleaved is GOAL.md's Challenge item, deliberately deferred — needs a design pass for non-API-harness resource contention).
• Multi-harness + replicate persists one report per runId (last harness wins) but displays all; single-harness is the primary path.
• Reconciliation: the K<5 warning fires only for 2 ≤ K < 5, so K=1 (single-run default) stays warning-free and unchanged.

Verification

• New tests: --replicate 5 → 5 runner calls/harness + report persisted & retrievable by runId; --replicate 1 unchanged (no report); K<1 errors; plus a JsonResultStore round-trip that survives a fresh store instance.
• Full unit suite 366/366; tsc --noEmit clean.

🤖 Generated with Claude Code

The merge-base changed after approval.