ENH: Rewrite core.prep to use vocles + vocalpy, remove io module

[NickleDave]

• use vocalpy classes to convert audio -> spectrograms if specified
• make a vocles dataset and save that by calling to_csv

if possible prefer clearly named functions, e.g. create_spectrograms_from_audio; current dataframe.from_files -> audio.to_spect -> spect.to_dataframe feels like concrete steps that can be written as statements but are obscured by calling out to multiple functions at different levels. It feels like there's 1 or 2 main pipelines we want to have built-in--spectrogram + audio datasets--and those should be written down and made explicit in code. Anything more exotic can be done by hand.