Use a tight bounding box when plotting at the intrinsic size

[seeM]

Addresses #5068. This brings our matplotlib configuration closer to Jupyter Notebooks when using the intrinsic sizing policy which is probably more familiar to users.

Since the produced image pixel size is no longer guaranteed, I also changed the save plot modal to display the size in intrinsic units instead of pixels.

Known Issues

If you plot totally outside of the default bounding box, it's missed by our current plot change detection and an "update" is not sent to the frontend. I was unable to find a fix in a reasonable amount of fiddling.

It's a bit of a hack but you can manually request an update with fig.canvas.manager.update(). Maybe a better workaround would be to add a refresh button to the plots action bar?

QA Notes

Running the following code in a single execution should include the title in the plot (you can't run it line-by-line because of the known issue above).

fig, ax = plt.subplots()
fig.text(x=0.5, y=1, s='title')
plt.show()

[timtmok]

I've wondered if we should add a refresh button. It does seem like a workaround if the plot needs it and the button bar is starting to get crowded. It could probably use a bit of a UI change to handle all these actions.

The changes look good! I think there might be an update needed to the automated tests but I don't think those tests do anything with intrinsic size.