Read MODULE.bazel files as UTF-8, and ignore decoding errors (#2981)

All Bazel files are now UTF-8 on all platforms, so this shouldn’t use
the platform-specific default encoding.

Since the content is only used to search for an ASCII string, we can be
lenient and ignore decoding errors (replacing invalid bytes with
U+FFFD).

See
https://github.com/phst/rules_elisp/pull/2178#issuecomment-6499182811.


### Issue description

See discussion in
https://github.com/phst/rules_elisp/pull/2178#issuecomment-6498656822

### How this PR solves it

Parse MODULE.bazel files as UTF-8 on all platforms, and ignore decoding
errors in this case.

### Testing

Waiting for CI
1 file changed