better waveform translation from mxcfb & implement WAVEFORM_MODE_AUTO

[raisjn]

I saw that for rM2, people switch from WAVEFORM_MODE_AUTO to WAVEFORM_MODE_DU - this is the correct thing to do, as WAVEFORM_MODE_AUTO is not implemented in rm2fb and is upgraded to GC16 (or GL16) and calls clearGhosting().

when using rm2, there is a different set of waveforms available: https://github.com/ddvk/remarkable2-framebuffer/blob/master/src/server/main.cpp#L51 and a translation happens from mxcfb_update -> the rM2 SWTCON API.

it's probably worth re-visiting and tuning the translation layer in rm2fb to get better behavior and maybe add a manual loop that picks the waveform mode to use if AUTO is specified and the region is small enough.

we don't actually have documentation on the rM2 modes and how many colors they support / their update timings for different region sizes. by using a calibration image, we can learn more about the modes exposed to us. (what does INIT support, for example)

(cc @mrichards42, @LinusCDE)