Allow internal audio processing with floats

[dagargo]

At the moment, it always works with short.

In particular, this is desired for audio transformations, like volume normalization, filtering...

But it is also possible that I'm overthinking all this.

[dagargo]

Also it is important when saving files using a different bit depth than the native one.

For instance, after doing some changes in a sample and saving it, if the file uses 24 bits, the resulting sample will still have 24 bits samples but as the editor uses only 16 bits there will be a loss of precision.