refactor(triggers): thin agent execution facade

[aaight]

Summary

Card: https://trello.com/c/69fdee572a6522696c72c6f1

This finishes the agent execution facade cleanup by keeping runAgentExecutionPipeline() as a high-level orchestration path and moving remaining runtime/follow-up details behind helper modules.

Changes

• Added agent-execution-runtime.ts for execution context setup, lifecycle hook loading, runAgent() invocation, post-run PR/work-item linking, PM summaries, and callbacks.
• Added agent-execution-followups.ts for recursive post-completion review dispatch and splitting auto-chain dispatch, while keeping recursion centralized through the facade.
• Removed the biome-ignore lint/complexity/noExcessiveCognitiveComplexity suppression from agent-execution.ts.
• Updated trigger docs to map the extracted helper modules and the current execution flow, including PR/work-item linking, PM summaries, post-completion review, splitting auto-chain, and auto-debug.

Verification

• npx vitest run --project unit-triggers --project unit-core tests/unit/triggers/shared/agent-execution.test.ts tests/unit/triggers/agent-execution.test.ts tests/unit/triggers/shared/webhook-execution.test.ts tests/unit/triggers/sentry-webhook-handler.test.ts tests/unit/pm/webhook-handler.test.ts
• npx vitest run --project unit-core tests/unit/architecture-docs.test.ts
• npm run lint:fix
• npm run lint (passes; reports existing complexity warnings in src/integrations/alerting/_shared/materialize.ts and src/triggers/sentry/webhook-handler.ts)
• npm run typecheck
• npm test

Integration tests were not run because this change does not alter handler wiring or credential scope behavior.

🕵️ codex · gpt-5.5 · run details

[codecov]

Codecov Report

❌ Patch coverage is 93.49112% with 11 lines in your changes missing coverage. Please review.

Files with missing lines	Patch %	Lines
src/triggers/shared/agent-execution-runtime.ts	88.88%	10 Missing ⚠️
src/triggers/shared/agent-execution-followups.ts	98.14%	0 Missing and 1 partial ⚠️

📢 Thoughts on this report? Let us know!

[nhopeatall]

LGTM - The facade remains source-compatible, the extracted runtime/follow-up helpers preserve the previous execution flow, and CI is green.

🕵️ codex · gpt-5.5 · run details