Add access to segment text as bytes

[travolin]

Access method needed for issues related to Issue #46

Change

This change adds an additional access method to the whisper_state to allow access to the segment text as a vector of u8.

Use Case

This is used to handle cases where the segment is not a valid utf-8 string and will throw an error when using full_get_segment_text. This can be pretty common for multi utf-8 byte languages like Mandarin. The segment is not guarantied to end on the proper utf-8 boundary, but can be found in the next segment. When an error is returned from full_get_segment_text the library user can simply grab the bytes (with this new method), buffer them and add the next segment to them until a proper utf8 string is found.

Testing

All testing was done against the following random Chinese language podcast "https://anchor.fm/s/1553388c/podcast/play/60927763/https%3A%2F%2Fd3ctxlq1ktw2nl.cloudfront.net%2Fstaging%2F2022-10-20%2F3b2b6ec4-4399-393e-c056-06e16f5c0c9e.mp3"

[tazz4843]

I was wondering what languages would throw this sort of issue, after mentioning it, now that makes sense. LGTM, thanks for the PR!