MRXS: conversion is incorrect when multiple pyramids are present

[melissalinkert]

Reported in https://forum.image.sc/t/issues-mrxs-to-tiff-using-bioformat2raw/100695; private sample data is in automated-tests/inbox/imagesc-100695/.

The problematic dataset has an extra Slide zoom mask level inserted between the Slide zoom level and Slide filter level (the "real" data). Looking at a diff of Slidedat.ini against the Slidedat.ini for https://openslide.cs.cmu.edu/download/openslide-testdata/Mirax/CMU-1.zip is useful.

melissalinkert@905daf4 is a start at fixing the channel count detection, but that does not result in correct image data. I suspect that pyramidDepth needs to be calculated and recorded separately for each hierarchy so that the index file is read properly, but more investigation is needed.