bugfix: convert non-`.wav`/`.raw` files to `.wav` before passing to pydub

[behackl]

Ongoing post-pyav refactor: fixes #3831

• [ ] cleanup inconsistently named containers/streams in added code
• [ ] think about actually opening file to check codec
• [ ] deal with files that have mulitple audio streams? (try whether this actually works currently)
• [ ] maybe add a test?