feat(agent): add Codex parser diagnostics to agent list

[codeaholicguy]

Summary

• Simplify agent list parser checks to focus only on true parse failures.
• Add a small adapter-neutral ParserHealthProvider / SessionParserHealth contract.
• Implement Codex parser health from the existing normalized session parser.
• Keep normal agent list output stable; warn only when a listed session has records but yields zero parsed messages.
• Keep agent list --debug on the existing debug logger pattern and emit only compact failure details.

Validation

• npm test --workspace @ai-devkit/agent-manager -- CodexAdapter.test.ts: passed, 68 tests.
• npm test --workspace ai-devkit -- agent.test.ts: passed, 76 tests.
• npm run typecheck --workspace @ai-devkit/agent-manager: passed.
• npm run lint --workspace @ai-devkit/agent-manager: passed.
• npm run lint --workspace ai-devkit: passed with existing unrelated warnings.
• npm run build --workspace @ai-devkit/agent-manager: passed.
• npx ai-devkit@latest lint: passed.
• npm run dev -- agent list --debug from packages/cli: inspected; emitted compact parser failures=0 debug output.
• Commit hook also ran repo-wide lint/test successfully.

Notes

• Unknown Codex record/payload types are no longer part of agent list output. The signal is intentionally narrower: session data exists but messages cannot be recovered.