fix(ci): repair merge detection and Rust Docker build

[konard]

Fixes #139

Summary

• Makes scripts/detect-code-changes.mjs event-aware: PR synthetic merge commits keep the existing PR-head diff behavior, while real merge commits pushed to main compare first parent to merge commit so the full merged branch is detected.
• Fixes the Rust Docker build by moving the builder image to rust:1.96-bullseye, satisfying the crate rust-version.
• Updates workflow policy to current template behavior: release-safe concurrency, actions/checkout@v6, and Rust actions/cache@v5.
• Moves repository-level CI-policy Jest tests into js/tests/ci/, excludes that directory from JS/Rust product parity matching, and keeps those test/config changes out of the blocking npm changeset gate while still running JS checks.
• Adds a case study with preserved CI logs, template diffs, root-cause notes, and verification logs under docs/case-studies/issue-139/.
• Reported the same detector defect upstream in the JS and Rust pipeline templates:
  • Detect real merge pushes with first-parent diff link-foundation/js-ai-driven-development-pipeline-template#91
  • Detect real merge pushes with first-parent diff link-foundation/rust-ai-driven-development-pipeline-template#81

Reproduction

• Failed Rust run 27504645835 logged Merge commit detected (pull_request event) during a push, emitted any-code-changed=false, then failed Docker build because rustc 1.94.1 did not satisfy rust-version = "1.96".
• JS run 27504645820 showed the same merge-event misclassification and false code-change outputs, though the release still completed.
• New Jest coverage reproduces the real merge-push false negative with a branch whose earlier commit changes Rust code and final commit is docs-only.
• Fresh PR parity run 27511525135 showed repository CI tests under js/tests/unit/ were being counted as JS product tests; the fix keeps those tests in CI while excluding js/tests/ci/ from product parity matching.
• Fresh JS run 27511646578 then showed the changeset gate counted js/tests/ci/ and js/jest.config.mjs as package code. js-code-changed now tracks publishable JS paths only: js/src/, js/bin/, and js/package.json.

Tests

• cd js && npm test -- --runTestsByPath tests/ci/detect-code-changes.test.js tests/ci/rust-dockerfile.test.js tests/ci/workflow-policy.test.js
• cd js && npm run lint passed with 0 errors and existing complexity warnings.
• cd js && npx prettier --check .
• GITHUB_BASE_SHA=$(git merge-base origin/main HEAD) GITHUB_HEAD_SHA=HEAD GITHUB_BASE_REF=main node scripts/check-js-rust-parity.mjs
• GITHUB_EVENT_NAME=push node scripts/detect-code-changes.mjs on the latest CI-test-only commit: js-code-changed=false, any-js-code-changed=true.
• cd rust && cargo fmt --all -- --check
• cd rust && cargo clippy --all-targets --all-features -- -D warnings
• cd rust && cargo test --all-features --verbose
• cd rust && docker build -t web-capture-rust-issue-139 .

Full local npm test was also attempted and is documented in the case study. It is not a clean local signal in this workspace because Playwright browser binaries are missing and the Docker e2e setup is normally prepared by CI.

[konard]

Working session summary

Implemented and pushed the fix to issue-139-24bf83def591.

PR 140 is ready: #140

Current head 6295928ecf9750b2c51ddac204df75a8b9c3991a is green:

• JavaScript Checks and Release: success
• Rust Checks and Release: success
• JS/Rust Parity: success

I also confirmed the PR is no longer draft and the working tree is clean. Older failed runs belong to superseded commits and were addressed by the latest push.

---

This summary was automatically extracted from the AI working session output.

[konard]

🤖 Solution Draft Log

This log file contains the complete execution trace of the AI solution draft process.

💰 Cost estimation:

• Model: GPT-5.5
• Provider: OpenAI
• Public pricing estimate: $32.068725

📊 Context and tokens usage:

GPT-5.5: (4 sub-sessions)

1. ~150K / 200K (75%) input tokens, 14.0K / 128K (11%) output tokens
2. ~150K / 200K (75%) input tokens, 14.0K / 128K (11%) output tokens
3. ~150K / 200K (75%) input tokens, 14.0K / 128K (11%) output tokens
4. ~383.9K / 200K (192%) input tokens, 35.8K / 128K (28%) output tokens

Sub-session values are estimates from observed compact events; the Total line remains exact.

Total: (833.9K + 20.2M cached) input tokens, 77.7K output tokens, $32.068725 cost

🤖 Models used:

• Tool: OpenAI Codex
• Requested: gpt-5.5
• Model: GPT-5.5 (gpt-5.5)

📎 Log file uploaded as Repository (2 chunks) (124336KB)

• View complete solution draft log

---

Now working session is ended, feel free to review and add any feedback on the solution draft.

[konard]

✅ Ready to merge

This pull request is now ready to be merged:

• All CI checks have passed
• No merge conflicts
• No pending changes

---

Monitored by hive-mind with --auto-restart-until-mergeable flag