Index data length have hard-coded length cap that only matches DAC paper parameters

The DAC paper cites the following parameters for acoustic index caluclations:

```
"--frame=16000"
"--hop=4000"
"--n_fft=1024"
```

For the dac dataset, which is split into 1 minute audio files, this results in acoustic indexes of length *either* 720 or 721, depending on the index.

The `SoundingOutDiurnal.to_dataframe()` function clips each index to a fixed, hardcoded length:

https://github.com/ecolistening/SoundADE/blob/ca33b6fcc5950b25abe3c5300a65f4263d5d32ac/src/soundade/data/datasets.py#L244

Where `n` is defined:

https://github.com/ecolistening/SoundADE/blob/ca33b6fcc5950b25abe3c5300a65f4263d5d32ac/src/soundade/data/datasets.py#L232

This clips each acoustic index to have length 720.

As long as the index calculation params are fine for the paper, then for the paper this does not *need* to change.

However for any other parameters this should be considered a bug, and the clipping shouldn't be hard coded, but derived from the params.