fix: restore batching for automatic finding matching

[ianw-oai]

Summary

Restore bounded batching for automatic post-scan history matching. The single-call restriction added in #638 can skip matching for a large history.

Changes

• Remove the allowBatching switch and its two single-call tests.
• Keep the existing batch size limit, group matching, and cancellation behavior.
• Correct the README's one-call claim and state where cost tracking stops.

Testing

• Focused matcher, component-scan, and API tests: 235 passed, 5 platform-specific tests skipped.
• pnpm run types, pnpm run format, and pnpm run build passed.
• git diff --check passed.

Risk and rollout

Automatic history matching can make several model calls. These calls remain outside the scan's recorded cost and maxCostUsd limit. This PR restores batching; it does not add post-scan budget accounting. No CLI flags or result formats change.

Public disclosure review

• No customer, partner, prospect, or user identities, data, or identifying details are included.
• No credentials, personal data, private source, scan findings, or nonpublic links or tickets are included.
• I reviewed the branch name, title, description, commits, changes, comments, logs, screenshots, attachments, and links for public disclosure.

[chatgpt-codex-connector]

Codex Review Summary

This comment shows the latest Codex review activity on this pull request.

Review	Status	Commit	Review trigger
🔒 Security Review	✅ Completed 2026-09-01T21:44:00.698558Z	c269641	PR opened

ℹ️ About Codex in GitHub

Your team has set up Codex to review pull requests in this repo. Reviews are triggered when you

• Open a pull request for review
• Mark a draft as ready
• Comment "@codex review" or "@codex security review".

Codex reacts with 👀 while any review is running, comments if it has suggestions, and reacts with 👍 once all reviews finish with no findings.

[ianw-oai]

Local validation for c269641 passed:

• Full suite in two orders (seeds 12345 and 2164852511): 2,208 passed, 43 skipped, no failures in each run.
• Focused matcher, component-scan, and API tests: 235 passed, 5 skipped.
• Type, format, build, and diff checks passed.

[sylvesterkaczmarek]

The batching path looks sound. Large comparisons are recursively split on one side, so every before/after pair is still covered once; each batch is validated before combineComparisons, then the combined result is validated again against the full input. Cancellation still happens between calls, and the README now makes the out-of-budget model calls explicit. I don't see a blocker here.