Add SRT (.srt) subtitle converter

[Bicheng (Kenneth) (BichengWang)]

SRT files currently go through the plain-text converter, so the output keeps cue numbers, millisecond timing arrows and styling tags, which is noisy for an LLM reader.

This adds SrtConverter, which turns each cue into one line prefixed with its start time:

[00:00:01] Hello there. Second line.

[00:01:05] Café time

• Cue numbers, end times, milliseconds and <i>, <b>, <u>, <font> and {\an8} styling are dropped. Other angle brackets (<Bob>, x<y) are kept.
• A cue runs from its timing line to the next one, so a blank line inside a cue or a missing separator between cues doesn't lose or merge text.
• Handles BOMs and CR, LF and CRLF line endings. If the detected charset fails on the full file (for example ascii guessed from the first 64 KiB), it falls back to a whole-file guess.
• A .srt with no cues raises, so the plain-text converter still handles it.
• Matches .srt, text/srt, text/x-srt and application/x-subrip.

No new dependencies. Tests are in tests/test_srt.py, and the full suite passes locally.

[Bicheng (Kenneth) (BichengWang)]

Hello members, is there any one can help additional review? Thank you!