Return deferred search content in Python SDK

[chruffins]

Summary

• generate the deferred search response model and return it from sync and async fetch calls
• update search content parameter and model descriptions from the current API schema

Testing

• uv run --with ruff ruff check on changed source and test files
• targeted pytest collection ran; all 25 mock-server cases are skipped by repository configuration

---

Note

Medium Risk
Breaking change for callers that assumed fetch returned None; behavior now depends on live deferred retrieval semantics and browser billing/concurrency documented in updated types.

Overview
Deferred search content fetch is now a typed API in the Python SDK: sync/async client.search.contents.fetch returns kernel.types.search.Response (with contents, search_id, usage, warnings) instead of None, and drops the old placeholder docs and Accept: */* / NoneType response handling.

A new search.Response model describes per-result Content entries (result_id, url, status, text, errors, etc.). Related OpenAPI-driven doc updates align inline and deferred content options (source, max_age_hours, browser/temporary-browser behavior) across content_fetch_params, search_create_params, result, and usage.

Tests and api.md now expect Response from fetch (including raw/streaming parse paths).

^{Reviewed by Cursor Bugbot for commit 5cf4120. Configure here.}

[chruffins]

Superseded by the Stainless-generated release PR #182, which includes this SDK update.