fix: Unable to route GET requests through proxy

[thiagobarbosa]

Fix #259

Summary

tiktoken.load.read_file (and the functions that call it) always used the module-level requests.get, so there was no way to route the download requests through a custom requests.Session. This adds an optional session parameter that flows through the whole loading chain, defaulting to None so existing behavior (requests.get) is unchanged when no session is provided.

Changes

• read_file(blobpath, session=None): uses (session or requests).get(blobpath) instead of always calling requests.get.
• read_file_cached(blobpath, expected_hash=None, session=None): accepts and forwards session to read_file.
• data_gym_to_mergeable_bpe_ranks(...): accepts session and forwards it to read_file_cached.
• load_tiktoken_bpe(tiktoken_bpe_file, expected_hash=None, session=None): accepts session and forwards it to read_file_cached.
• Added tests/test_load.py with a fake requests.Session to verify read_file uses the provided session instead of the default requests module.

Test plan

• pytest tests/test_load.py — new test verifies a custom session is used for the HTTP GET instead of the module-level requests.get.
• Existing behavior preserved: calling read_file/read_file_cached/load_tiktoken_bpe/data_gym_to_mergeable_bpe_ranks without a session still uses requests.get directly.