Clarify or handle non-Latin-1 str input in parse_options_header

[shmulc8]

parse_options_header accepts str | bytes | None, but passing a str
containing characters outside Latin-1 raises UnicodeEncodeError.

Reproduction

from python_multipart.multipart import parse_options_header

parse_options_header(
    'form-data; name="upload"; filename="中文.doc"'
)

Observed on python-multipart 0.0.32 and current main
(d9cb4c62db5b9defeeaa99bfe0e4da51e760108f) with Python 3.14.6:

UnicodeEncodeError: 'latin-1' codec can't encode characters in position 0-1

This does not appear to affect ordinary HTTP multipart processing: raw header
bytes round-trip correctly, and Unicode filenames work through Starlette.
This is an API-contract/robustness question, not a security report.

Would you prefer to:

1. Document that str inputs must be Latin-1/WSGI-compatible;
2. Validate this explicitly and raise a clearer error; or
3. Support arbitrary Unicode str values?

I would be happy to submit a focused PR with tests once the intended behavior
is clear.