[DylanCosto]

Genome-wide robustness experiments can run for many minutes with no intermediate output. Add readable stage changes and periodic term counts on stderr, with --progress json for workflow tools, --progress quiet, and a positive --progress-interval (30 seconds by default).

Updates distinguish processed, common-callable baseline and excluded terms for each score. Scoring checks the interval between blocks of 256 terms; this is not a heartbeat during input loading or a runtime estimate. The final complete event is emitted only after input re-verification and publication of report.json. Progress settings and timestamps stay outside scientific outputs. Ordinary error diagnostics still appear on stderr, and a closed log consumer does not invalidate the experiment.

Validation:

• Full stable suite: 182 tests passed, one optional local-data test ignored; formatting and all-target Clippy passed. Both CLI regressions also passed on Rust 1.88.
• CLI checks cover identical scientific output bytes in all three modes, exclusion accounting, rejected zero intervals and no success event after a failed baseline scan.
• Full-model public pilot: 35 scoring events over 1,745,179 terms; 1,712,283 retained baseline terms and 32,896 exclusions reconcile. All four numerical payloads and scenario summaries match the previously checked pilot.
• Extracted package: 52 payload hashes, 50 documentation links and the offline paired study passed.

Stacked on #4. The prepared 128-person release candidate is preserved; this follow-on does not change its evidence or publish a release.