feat(ci): enable AI review commands and activity events

[zhongkechen]

Summary

• mirror the pinned reusable AI review workflow's supported event triggers
• enable authorized /ai review commands through pull request issue comments
• record activity for submitted, edited, or dismissed reviews and created, edited, or deleted review comments
• add regression coverage for the caller event contract

Follow-up to #695.

Validation

• parsed the modified workflow with PyYAML
• ran both AI review workflow regression assertions directly
• ran git diff --check
• ran .github/scripts/lintcommit.py

[github-actions]

Codex AI review

No actionable findings. Residual risk is limited to live GitHub event/reusable-workflow integration, which the text-based regression test does not exercise.

Reviewed commit 8596f320dfea932f036536332d8910516e090f61. Workflow run

[github-actions]

Claude AI review

This PR only touches two files: the ai-pr-review.yml reusable-workflow caller and its regression test. It adds pull_request_review, pull_request_review_comment, and issue_comment to the existing pull_request_target trigger set, matching the description ("mirror the pinned reusable AI review workflow's supported event triggers").

Verification performed:

• Confirmed the added on: block in .github/workflows/ai-pr-review.yml exactly matches the literal string asserted by the new test_ai_pr_review_caller_subscribes_to_supported_events test (byte-for-byte, including indentation and event types: lists), so the test will pass and does exercise the actual change.
• The event types chosen for each trigger (submitted, edited, dismissed for pull_request_review; created, edited, deleted for pull_request_review_comment; created only for issue_comment) are valid, complete GitHub Actions type enumerations for those events.
• No changes were made to permissions:, secrets:, or the with: inputs passed to the reusable workflow, so the caller's authorization/permission surface is unchanged apart from the new event types that can now reach it.
• The new test follows the existing brittle-but-established convention (plain substring match on the raw YAML text) already used by test_ai_pr_review_caller_grants_reusable_workflow_permissions, so it is consistent with house style rather than a new regression.
• No core SDK, testing, OTel, or examples packages are touched, so determinism/replay, serialization, and public-API concerns from AGENTS.md do not apply to this change.

No confirmed defects found in the diffed lines.

Residual risk (not blocking, for awareness): correctness of the new triggers ultimately depends on the pinned reusable workflow (aws/aws-durable-execution-ci/.github/workflows/ai-pr-review.yml@8de63fa6...) correctly scoping issue_comment handling to pull-request comments and enforcing commenter authorization before acting — that logic lives outside this repository and isn't exercised by any test here. If that downstream authorization is ever weakened, this caller would now deliver issue_comment/pull_request_review/pull_request_review_comment events (including on non-PR issues, for issue_comment) into a job that still carries contents: write, pull-requests: write, id-token: write, and secrets: inherit. Also, the new regression test validates the change only via literal text containment rather than parsing the YAML on: mapping, so an equivalent but differently formatted trigger block would fail the test even though it were functionally correct.

Reviewed commit 8596f320dfea932f036536332d8910516e090f61. Workflow run