Fix nonexistent iter_content() reference in stream() docstring

[cognis-digital]

The docstring for the public httpcore.stream() function tells users how to
read a streamed response body:

If you want to access the response body you should either use
content = response.read(), or for chunk in response.iter_content().

But Response has no iter_content() method — the streaming iterator is
iter_stream() (and aiter_stream() for async). Running the documented
for chunk in response.iter_content() raises AttributeError.

The rest of the project already uses the correct name (docs/quickstart.md
documents for chunk in response.iter_stream():, and the tests use
response.iter_stream()). This one docstring drifted because doctests are
not run in CI.

This corrects the docstring to iter_stream() and adds a small regression
test pinning the documented streaming method names, so the documented API
can't silently drift to a nonexistent method again.

No library behaviour changes — docs-only correctness plus a guard test.

[fallenmi]

Reviewed with Codex.

I checked this on exact head 08901009255b95b7500f2a19f6695f9983e581ae. The exact base/current master (10a658221deb38a4c5b16db55ab554b0bf731707) directs users to Response.iter_content(), which is not implemented, while Response.iter_stream() is the actual synchronous streaming API. On the head, the docstring and public API now agree.

Validation:

• repeated base/head doc-to-API oracle 100x (base mismatch, head match)
• pytest -q tests/test_models.py::test_response_streaming_api_names -> 1 passed
• focused Ruff format/check and mypy -> clean
• git diff --check -> clean
• live matrix: 7/7 successful

No blocking issues found.