chore: add AGENTS.md and CLAUDE.md

[sai-ray]

This PR:

• Adds AGENTS.md at the repo root: an agent-facing repository guide auto-discovered by Codex CLI, Copilot, Cursor, Claude Code, Kiro, and other coding tools.
• Adds CLAUDE.md: a one-line @AGENTS.md import so Claude Code picks up the same content deterministically.
• No existing docs are modified.

Why

AI coding agents working on this repo today have no auto-discoverable entry point. Agents either skim CONTRIBUTING.md (human-oriented prose) or operate without project-specific guidance, leading to inconsistent behavior across tools (jsii incompatibilities, edits to projen-generated files,
missing alarmNames, snapshot drift mistakes).

AGENTS.md is the cross-tool industry standard (governed by the Linux Foundation's Agentic AI Foundation; ~60K repos as of March 2026). The same pattern is being adopted in aws/aws-cdk and across the cdklabs org.

What

• AGENTS.md (~270 lines) — quick reference, codebase non-obvious locations (with lib/ and projen-generated siblings flagged "NEVER edit"), role & principles, architecture, MUST/SHOULD/MAY design rules (jsii, projen, file naming, tests, alarms, build hygiene), decision logic for common
  branching choices (Lambda vs ECS vs shared, alarm severity, public type placement, integ test scope, AlarmSeverities vs alarmOverrides), implementation patterns, contribution process, key references.
• CLAUDE.md (1 line) — @AGENTS.md import.
• Content is extracted from CONTRIBUTING.md, README.md, docs/application-overview.md, .projenrc.ts, and the conventions already enforced in src/. No new rules are introduced.
• Severity assignments (MUST/SHOULD/MAY) reflect what existing code and CI already enforce. Worth a sanity check on the alarm and projen sections in particular.

---

By submitting this pull request, I confirm that my contribution is made under the terms of the Apache-2.0 license