fix: add explicit encoding="utf-8" to .txt read in _validators.py

[Ghraven]

What

src/openai/lib/_validators.py reads a user-supplied .txt fine-tuning file in text mode without an explicit encoding:

with open(fname, "r") as f:
    content = f.read()

In text mode, open() uses the platform default encoding (locale.getpreferredencoding()). On Windows that is typically cp1252, not UTF-8.

Why it matters

.txt datasets frequently contain non-ASCII characters (smart quotes, accented text, CJK, emoji). On a default-cp1252 platform, reading a valid UTF-8 file raises UnicodeDecodeError or silently corrupts characters.

Before / After

# Before - platform-dependent
with open(fname, "r") as f:
# After - consistent across platforms
with open(fname, "r", encoding="utf-8") as f:

How I verified

Reproduced the failure on a simulated cp1252 default: the explicit utf-8 read preserves content, while a cp1252 read of the same valid UTF-8 file raises UnicodeDecodeError. With encoding="utf-8" the read succeeds regardless of platform. No behavior change on systems that already default to UTF-8.

Scope

Single-line change in src/openai/lib/, a hand-maintained (non-generated) path per CONTRIBUTING. Happy to adjust if you would prefer a different approach.