Refactor plotters

[oerc0122]

Description of work

• Refactor all the plotters to use generator constructs, gridspecs and attempt to reduce code-duplication.
• Adds new "grouped" plot type with one plot per dataset.

Fixes
Prevent number of plots corresponding to number of datasets.

To test
Try to break plotting with weird sequences of button presses.

[MBartkowiakSTFC]

I have been testing the plotter. What I found so far is:

1. Grid plots of 1D datasets have an empty legend.
2. If I plot into Grid or Single with legend switched off, and switch it on afterwards, I get an exception: <class 'AttributeError'> 'NoneType' object has no attribute 'set_visible'
3. Heatmap plot is labelling the plot axes wrong. In my test, time was shown as Q and Q was shown as time.

4. Heatmap plots start off without any legend. If I switch legend off and on, the legend appears.
5. If I try to plot a vector summary from the plot creator like this

the plot type ends up being Single and not Vectors:

While we do not want to allow Vectors plot type for normal data sets, it should still be possible to use Vectors for vectors.

6. For Single and Heatmap plots, switching the legend off and on makes all the subplots resize.
   Before:

After:

[MBartkowiakSTFC]

I spent some time plotting data using this branch. A lot of it works already.
I think that these points would need to be looked at:

1. I managed to make the Heatmap plot fail by changing the main axis of one of the 3 subplots:

Similarly, unchecking the "Use it" checkbox for one of the datasets makes the entire `Heatmap` (or `Grouped`) plot fail.

2. When I got the Heatmap plot of DISF to appear using the Q vector as the main axis, the plot appears reversed. The signal should become broader in energy at large Q values, but here it is the other way around:

3. In Single plots, the legend for each curve used to include information on the position of each curve on the other axis. That is, this DISF plot would normally have a legend telling me that these curves correspond to different Q values:

`Grid` and `Grouped` plots in this branch do it correctly and show the q position of each cut.

4. After switching the main axis there and back (to Q and to energy again) and going through Single, Grid, Grouped plotters, when I go back to Heatmap I get a plot where both axes are the energy axis:

[MBartkowiakSTFC]

I think it would make sense to merge this PR once the linter is happy with it.

Normalisation of heatmap plots will require more work and will be best dealt with in a separate PR.
The only other detail is that the first plot (available vs. selected vectors per shell) in vector statistics is now a histogram (and it used to show a point per shell). However, the values are still correct. We can always decide at a later time which plotting style works the best.

[ChiCheng45]

Some issues with the plotter, 3D plotter, and resizing.

[MBartkowiakSTFC]

Some issues with the plotter, 3D plotter, and resizing.

OK, you're right. It can happen on my computer too, if I shrink the GUI window enough.

In case it helps, my first guess is that it has something to do with the line self._figure.set_layout_engine("tight") which has been commented out in this PR.

[oerc0122]

This issue is actually due to the new length of the legend. I am going to also make long legend labels collapse with a ... which seems reasonable.

[MBartkowiakSTFC]

This issue is actually due to the new length of the legend. I am going to also make long legend labels collapse with a ... which seems reasonable.

This may be the case. Still, you can create a legend larger than the plot itself also in the protos branch, and there the plot does not get resized.

On my machine I found that the minimal change to this branch that eliminates this specific problem is to remove the layout keyword argument from PlotWidget line 384
mpl.figure(layout="constrained")
and just create a default figure instead
mpl.figure()
But this leads to other problems later when switching to Heatmap and back, since Heatmap changes the matplotlib layout engine at the moment. There could be other consequences in other plot types too...

[ChiCheng45]

On the single plot, the number of plots is limited to a small number of line plots per dataset. Can you also do this from the grouped plots? There are some issues if you have a lot of line plots and you use the Y offset.

Here I have 441 line plots. When Y offset is used, it becomes quite messy.

Also, on the grouped plot, the home button doesn't work as expected compared to single. On single plot, the home button resets the plot so it fills the space; this doesn't always happen on grouped.