ak4619/cal: test sample widths other than W=16

[vk2seb]

Initially raised in #37

There's nothing limiting us to W=16-bit samples (maximum from AK4619 is 32 bits per sample).

We should try some different sample widths and make sure the rest of the modules (calibrator, demos) all still work.

The calibrator module cal.sv would need to be modified to be parameterized across different fixed-point representations.

[vk2seb]

This is tested now at W=8 and W=32 with some caveats listed here: #48

Still need to fix up the cal script <-> cal.sv interaction for the DC zeroing process. But at least it is possible to build the mirror core and have different sample widths pass through the chain correctly.