fix(agent-scan): honor provider timeout_ms in HTTP requests

[rwinkelman]

Fixes #613

Summary

Honor timeout_ms from provider config in agent-scan HTTP requests.

Test plan

• Adapter passes timeout to HTTP client

[boy-hack]

Thanks @rwinkelman for fixing #613. Honoring timeout_ms at the httpx client level is the right place, and threading it through all four provider call paths (_call_standard_provider, _call_http_provider, _call_dify_provider, _call_coze_provider) is consistent. _get_timeout_seconds correctly falls back to self.timeout when config is None, so the if config else None guards are correct.

One heads-up before merge: this PR overlaps heavily with the maintainer's in-flight PR #617, which edits the exact same _make_http_request signature and the same four call sites. The only difference is the parameter name — this PR uses timeout=..., while #617 uses timeout_seconds=... (with a richer docstring). Merging both will produce a conflict on adapter.py.

Suggestion:

• Coordinate with fix(agent-scan): honor provider-configured timeout_ms for HTTP targets #617: pick one parameter name (timeout_seconds reads clearer as a keyword) and rebase this branch on top of fix(agent-scan): honor provider-configured timeout_ms for HTTP targets #617, or let fix(agent-scan): honor provider-configured timeout_ms for HTTP targets #617 absorb the change.
• Minor: effective_timeout in _make_http_request is a good rename — consider keeping it regardless of which PR lands first.

Otherwise the logic is sound. Once the conflict with #617 is resolved, this is good to merge.

[rwinkelman]

Grooming pass: timeout_ms honor for HTTP provider. Ping if anything else is needed before merge.

[boy-hack]

Thanks for the grooming pass, @rwinkelman — the change is still sound. Just a reminder that the merge conflict with #617 stands: my in-flight PR #617 edits the exact same _make_http_request signature and the same four call sites, only with parameter name timeout_seconds (and a richer docstring) instead of timeout. Until one of them lands, they can't both be merged cleanly.

Since #617 is mine, I'd suggest one of:

• I let fix(agent-scan): honor provider-configured timeout_ms for HTTP targets #617 absorb this (it already threads the provider timeout_ms through all four paths with the effective_timeout rename you also liked), then close fix(agent-scan): honor provider timeout_ms in HTTP requests #625 as superseded — least churn.
• Or you rebase fix(agent-scan): honor provider timeout_ms in HTTP requests #625 onto fix(agent-scan): honor provider-configured timeout_ms for HTTP targets #617 and we keep your timeout naming; I'd then update fix(agent-scan): honor provider-configured timeout_ms for HTTP targets #617 to match.

Either way the logic is correct and checks are green. Want to coordinate on which one we keep?