Advanced diplotype clustering - investigate CNV imshow weirdness

[alimanfoo]

Hi @sanjaynagi, I did some digging into the CNV imshow weirdness. In this PR I stripped back the _plot_dendro_cnv_bar() method to the simplest possible implementation. I then tested it in the spike_dipclust_cnv notebook, outputting both the final figure and the CNV subplot.

What's weird is that the CNV subplot respects the color_continuous_scale parameter and the zmin and zmax parameters, but in the final combined figure these are ignored, meaning that the default PuOr colorscale is always used, and the zmin and zmax are always obtained from the data.

I think this has to be something to do with the way that the subplots are combined into a single figure within the concat_subplots() method. Something about this is erasing some of the parameters originally given to imshow.

That's as far as I've got for now, hope it's useful :)

[alimanfoo]

I did a bit more digging, have a basic working version here. Looks like we have to use go.Heatmap() rather than px.imshow() otherwise the zmin, zmax and colorscale parameters get lost.