fix(python): add encoding=utf-8 to file reads and writes in sandbox.py

[mesutoezdil]

read_text() and fdopen(..., "w") without an explicit encoding use the system locale,
which can be non-UTF-8 on some platforms (e.g. Windows).

4 call sites in sandbox.py read or write UTF-8 config files without specifying encoding:

• metadata_path.read_text() gateway metadata.json
• token_path.read_text() oidc_token.json
• os.fdopen(fd, "w") atomic write of oidc_token.json
• active_file.read_text() active_gateway name file

All 4 are now explicit with encoding="utf-8".

[copy-pr-bot]

This pull request requires additional validation before any workflows can run on NVIDIA's runners.

Pull request vetters can view their responsibilities here.

Contributors can view more details about this message here.

[johntmyers]

gator-agent

PR Review Status

Validation: This is a small, concentrated Python SDK bug fix that makes OpenShell's shared gateway config files use an explicit UTF-8 encoding, matching the Rust-side writers for metadata.json, active_gateway, and oidc_token.json.
Head SHA: dba017a42c918bbda280010820f75f0ab4d0811b

Review findings:

• python/openshell/sandbox.py: _read_oidc_token_bundle() now decodes with explicit UTF-8, but UnicodeDecodeError is not caught. The function contract says unreadable token files return None, so a corrupt/non-UTF-8 oidc_token.json should fail closed like malformed JSON:

  except (OSError, UnicodeDecodeError, json.JSONDecodeError):
      return None

• Please add focused regression coverage for the locale/encoding behavior. Existing tests mostly use locale-default write_text() / read_text(), so they would not catch a future removal of encoding="utf-8" on non-UTF-8 platforms. A small test that writes non-ASCII UTF-8 bytes directly for metadata.json, active_gateway, and oidc_token.json would cover the important paths.

Docs: not needed; this does not change a documented command, config field, or workflow.

Next state: gator:in-review

[johntmyers]

/ok to test dba017a