docs: correct create_mcp_http_client default timeout docstring

[aryanmotgi]

Summary

The docstring for create_mcp_http_client in src/mcp/shared/_httpx_utils.py claims the default timeout is "30 seconds if not specified". However, the implementation sets:

kwargs[\"timeout\"] = httpx.Timeout(MCP_DEFAULT_TIMEOUT, read=MCP_DEFAULT_SSE_READ_TIMEOUT)
# i.e. httpx.Timeout(30.0, read=300.0)

So the read timeout is actually 5 minutes (to accommodate long-lived SSE streams), not 30 seconds. This is an inaccuracy users will hit if they rely on the docstring to reason about long-running requests.

This PR updates the two affected lines in the docstring to reflect what the code actually does. No behavior change.

Test plan

• Docstring-only change; no functional impact
• Matches the real defaults (MCP_DEFAULT_TIMEOUT = 30.0, MCP_DEFAULT_SSE_READ_TIMEOUT = 300.0) defined at the top of the same file

[Shivansh12t]

Good catch, the previous wording implied a flat 30s timeout, which would be surprising to anyone debugging a dropped SSE connection or trying to understand why httpx.Timeout(30) behaves differently from the default. The split between connect/write/pool (30s) and read (300s) is the actual contract here, so surfacing it in the docstring is worth it.
One minor suggestion: the inline comment and the Args: block now say essentially the same thing in slightly different phrasing. Could consolidate to just the Args: section and drop the duplicate from the overview bullet, or make the bullet a one-liner that points to the rationale (SSE streams) without restating the numbers.

[aryanmotgi]

Good call — consolidated the overview bullet to a one-liner that points to the Args section, which now carries the actual numbers and rationale. Pushed in 452b15e.