Add execution latency observations and trace correlation

[sam2tom]

Cold input submission can wait many seconds before executor preparation, while existing logs measure only control readiness and Core-visible first text. Add stage timings and identity bridges so admission validation/ownership, reservation waiting, worker selection, Runtime lifecycle gates, provider operations, connection observations, Environment initialization and executor preparation can be distinguished.

Link the submitting HTTP trace to a diagnostic execution trace derived from the durable reservation identity. Carry that trace through the existing Core–daemon Envelope.Trace field, preserve context in readiness/start logs, and distinguish preparation_request_id from the HTTP request_id. Session creation with initial inputs links its HTTP origin through Session identity to the worker's initial reservation.

The operator guide documents overlap, queue-age clock skew, background correlation and unavailable model-only TTFT. Scheduling, admission, cancellation, retries, native packages and billing behavior are unchanged. No SQL migrations, database columns, wire fields or deployment configuration are added. This PR targets beta because production currently runs its code path; it does not deploy or activate a template.

Validation: execution package regression and race tests; gateway/log package regression; make check-runtime-contract; make check-docs; make check-ci (47 tests); name guard (15 tests); translation source/hash checks; diff checks; two fresh independent reviews. Website CI passed, including the complete translation-structure check. No additional paid model calls or production mutations.

[chatgpt-codex-connector]

You have reached your Codex usage limits for code reviews. You can see your limits in the Codex usage dashboard.
To continue using code reviews, add credits to your account and enable them for code reviews in your settings.