loading only necessary audio data in memory

[LoannPeurey]

Is your feature request related to a problem? Please describe.
Some functions load audio data for analysis (zooniverse extract, derivation of acoustics). They load the entire file and then analyze the relevant segments. This takes significant time on large audios, and can lead to memory issues.

Describe the solution you'd like
Rather than loading full audio, load only the vocalization audio, this would work best with threading

[LoannPeurey]

Trying librosa load on only small sample of audio for acoustic derivations fro Philippines dataset.
Should get a better view of effectiveness after this

[LoannPeurey]

• Acoustic derivation
• zooniverse

quick and dirty benmark : 314s for Acoustic, run old to compare :