M15 candidate: harden conversation engine docs and conformance

[roackb2]

Summary

• add product-facing docs for the alpha conversation engine vs runAgentLoop layers
• add a realistic conversation engine example and programmatic guide fixes
• extend import-boundary conformance checks for deleted chat paths and obsolete names

Verification

• yarn build
• yarn test
• git diff --check
• yarn vitest run src/tests/unit/core/import-boundaries.test.ts src/tests/unit/core/conversation-engine.test.ts
• rg -n "core/chat/session-submit|core/chat/trace|core/chat/storage|core/chat/session-lease|core/chat/session-turn-result|core/chat/turn-persistence|core/chat/conversation-turn|core/chat/compaction|core/chat/turn-host|core/chat/turn-host-bridge|submitChatSessionPrompt|executeOrdinaryChatTurn|clearOrdinaryChatTurnLease" src docs README.md examples

Notes

• candidate implementation only; not claiming M15 complete
• scope kept to docs, examples, and conformance tests only

[chatgpt-codex-connector]

💡 Codex Review

Here are some automated review suggestions for this pull request.

Reviewed commit: c26ed6c579

ℹ️ About Codex in GitHub

Your team has set up Codex to review pull requests in this repo. Reviews are triggered when you

• Open a pull request for review
• Mark a draft as ready
• Comment "@codex review".

If Codex has suggestions, it will comment; otherwise it will react with 👍.

Codex can also answer questions or update the PR. Try commenting "@codex address that feedback".

[chatgpt-codex-connector]

Resolve relative import paths before removed-path checks

The new removed-path conformance test only catches specifiers that already start with core/chat/..., so it misses reintroductions via relative imports. For example, a production file under src/core/chat/engine/... could import a deleted module as ../trace.js; normalizeImportSpecifier leaves that as ../trace, which will never match prefixes like core/chat/trace, and the test still passes. This weakens the regression guard the commit is adding for deleted chat paths.

Useful? React with 👍 / 👎.