[TASK] Check the sample rate interacts properly with `n_fft`, `hop_length` and `frame_length`

[nrs23]

Do we need these three variables to scale with sample rate?

Currently they are fixed by command line arguments, regardless of sample rate of the audio file, which could vary within the same dataset - possibly resulting in inconsistent analyses?

[m4gpi]

Yes, if you want to preserve temporal spacing, these should scale according to the sample rate. For example, if our sample rate is 48K, and we set a n_fft to 2048, if we down-sample to 24K, we should use n_fft=1024.

However, I don't think its the job of the soundade to ensure this unless users were asked to specify values in seconds, but that could cause problems since they'll have to specify floating point values and it might not be possible to resolve fixed integer values for those parameters.