fix(tests): make the test suite pass on a default Windows checkout

[MohammedAlkindi]

• I understand that this repository is auto-generated and my pull request may not be merged

Changes being requested

On a fresh clone on Windows with stock Git settings, pytest fails 4 tests before any code is touched:

FAILED tests/test_client.py::TestOpenAI::test_proxy_environment_variables
FAILED tests/test_client.py::TestAsyncOpenAI::test_proxy_environment_variables
FAILED tests/test_transform.py::test_base64_file_input[sync]
FAILED tests/test_transform.py::test_base64_file_input[async]
4 failed, 7120 passed, 29 skipped in 71.94s (0:01:11)

Two unrelated mechanisms, one tiny fix each:

1. test_base64_file_input asserts on the fixture's exact bytes (SGVsbG8sIHdvcmxkIQo=, i.e. "Hello, world!\n"). Git for Windows defaults to core.autocrlf=true, so checkout rewrites tests/sample_file.txt to CRLF and the encoding becomes SGVsbG8sIHdvcmxkIQ0K ("Hello, world!\r\n"). The repo has no .gitattributes, so nothing pins the fixture's bytes — added one marking that single file -text. The committed blob is already LF, so no renormalization is involved; only future checkouts change.

2. test_proxy_environment_variables (sync and async) sets HTTPS_PROXY first and then deletes the lowercase variants as cleanup. os.environ is case-insensitive on Windows, so monkeypatch.delenv("https_proxy") also removes the HTTPS_PROXY that was just set — the client sees no proxy in the environment and the mounts assertion fails with len(mounts) == 0. Reordered: cleanup first, then setenv. On POSIX the two orders are equivalent (the lowercase names are genuinely distinct variables there), so nothing changes for existing platforms.

After the two changes, same machine:

tests/test_client.py::TestOpenAI::test_proxy_environment_variables PASSED
tests/test_client.py::TestAsyncOpenAI::test_proxy_environment_variables PASSED
tests/test_transform.py::test_base64_file_input[sync] PASSED
tests/test_transform.py::test_base64_file_input[async] PASSED

and both files in full: 254 passed, 2 skipped (the skips are pre-existing). I re-ran the affected files, not the whole 7k-test suite, for the "after" numbers.

Additional context & links

• Neither edited file carries the Castiron generated-file banner, and .gitattributes is new, so nothing generated was touched. If these tests are actually maintained in your generator templates, the two changes are deliberately minimal so they should be easy to port there instead.
• Environment: Windows 11, git 2.53.0.windows.2 (default core.autocrlf=true), Python 3.12.10, dependencies from requirements-dev.lock, mock server started per scripts/mock (@stdy/cli@0.22.1).