Plot interface cannot support line-profile datasets

[matbryan52]

Came across this when working with some user data:

  File "C:\ProgramData\Miniconda3\envs\GMS_LIBERTEM\lib\site-packages\libertem\viz\base.py", line 158, in __init__
    channel, eligible_channels
ValueError: channel intensity not found or not plottable, have: []

as the dataset has a shape (1, 600, sy, sx). The issue comes from our filtering of plottable buffers to those with a 2D shape after a .squeeze.

LiberTEM/src/libertem/viz/base.py

Lines 74 to 80 in da4230c

	def get_plottable_2D_channels(buffers):
	return [
	k
	for k, buf in buffers.items()
	# 2D data, removing axes of size 1
	if len(buf.data.squeeze().shape) == 2
	]

I think we could afford to be more selective in this function, for example 2D nav or sig buffers with no extra_shape or a 'squeezeable' extra shape can remain 2D even if any of their first two dimensions are 1.

Of course it's not necessarily useful to plot a (1, 600) image, but supporting it would prevent the need for special-casing in user code unless that's desired (to get a graph-style plot, for example)!

[matbryan52]

Would also need to supported in the default extract, of course.