Add PDF upload + Responses streaming example and ensure MIME via filename

[oliviergodement]

• Problem: Uploading raw PDF bytes without a filename leads to the server receiving an unsupported MIME type (None), causing 400 errors when referenced in Responses streaming. Also, the correct content payload for Responses should use input_text + input_file. See Issue #2472.

• What this changes

  • Example: Adds examples/responses/file_input_pdf_streaming.py showing:
    • Setting a filename on an in-memory buffer (so the server infers application/pdf), or
    • Using tuple-style upload (filename, bytes, "application/pdf").
    • Using input=[{"role": "user", "content": [{"type":"input_text"}, {"type":"input_file", "file_id": ...}]}] with client.responses.stream(...).
  • Tests:
    • tests/test_multipart_pdf_upload.py: Asserts multipart includes filename="document.pdf" and, when tuple is used, Content-Type: application/pdf.
    • tests/api_resources/test_responses_file_input.py: Mocks file upload + /responses and verifies a streaming call with input_file completes.

• Why this works

  • Providing a filename (or tuple with explicit content type) ensures correct MIME handling on the server.
  • Using input_file with input_text matches the Responses API input schema and avoids malformed content arrays.

• How to run

  • Example:

    python examples/responses/file_input_pdf_streaming.py /path/to/document.pdf --prompt "Summarize this document."

  • Tests:

    pytest -k "multipart_pdf_upload or responses_file_input"

• Backward compatibility

  • No breaking changes; adds example and tests only.

• Closes

  • Fixes #2472