[MSC3246](https://github.com/matrix-org/matrix-spec-proposals/pull/3246) specifies the waveform as a list of ints because:
> Because floating point numbers are not allowed in Matrix events
Though DSP on audio data is almost always done using their floating point representation.
This PR brings the float<->int rescaling in the `matrix` module so that the application code can always work with float waveform samples.
0e4141863b