Add Chatgpt Batch Detail Fetch

The ChatGPT connector currently hydrates conversation detail with one GET /conversation/{id} request per conversation. Large accounts turn that into a provider request storm, increasing run time and 429 pressure even though ChatGPT exposes a batch detail endpoint that returns the same conversation-detail shape for up to 10 ids per request.

The existing ChatGPT connector already has the core correctness machinery this change needs: list pagination, detail parsing, DETAIL_GAP handling, retry/backoff, source-pressure handling, run budgets, and convergence tests. The problem is the number of detail requests, not emitted record semantics.

The ChatGPT connector SHALL hydrate listed conversation detail by using the provider batch conversation-detail endpoint when available. A batch request SHALL include at most 10 conversation ids. The connector SHALL preserve the existing per-id GET /conversation/{id} detail path as a fallback for ids omitted from the batch response and for batch endpoint unavailability.