feat: MatPlot updates (Updated version of #337)

[nulinspiratie]

Fixes #257, and replaces #337 since that branch was deleted. Also includes fixes proposed in #615.

Changes proposed in this pull request:

• Interpolation of x and y coords in pcolormesh. Previously, MatPlot used to ignore edgepoints.This is because MatPlot treats the x and y values as grid corners instead of grid centers. This also resulted in data points being shifted by half a grid. Now the x and y values are transformed to grid corners (which increases their arrays by 1). Doing this allows the edgepoints to be included as well. If the array only contains one element, the grid size cannot yet be determined, and defaults to 1. This also works with non-equidistant setpoints.
• Args passed in initial instance creation are treated as data that should be plotted on individual subplots. This means that MatPlot(data.arr1, data.arr2) will create two subplots, one for each arr. An arg can also consist of multiple args, in which case they will all be plotted on the same subplot (e.g. MatPlot([data.arr1, data.arr2], data.arr3)` will create two subplots: one for arr1 and arr2, the other for arr3).
• Easy access to subplots via subplot index (plot[k] is now identical to plot.subplots[k]). The add method is furthermore added to each subplot, meaning that you can now write plot[k].add(data), which is identical to plot.add(data, subplot=k). A consequence is that the method _get_axes is not needed anymore, and is therefore removed.
• Inferring subplot shape from args unless explicitly provided. MatPlot now has a class attribute max_subplot_columns (=3 by default), which indicates the default maximum number of subplot columns. If one passes 3 args and no subplots kwarg, it will use subplots=(1,3) by default. For 6 args, it will become subplots=(2,3).
• Inferring figsize from subplots unless explicitly provided. The method default_figsize() returns a default figsize to use depending on the shape of subplots.
• Scaling of colorbar in pcolormesh
• Add method tight_layout(), which performs fig.tight_layout() with some extra space at the top for the title. It is called at the end of initialization

@giulioungaretti @jenshnielsen

[jenshnielsen]

The default has changed from 1 to 0 and this is now zero based. Is that intentional?

[nulinspiratie]

Ah sorry I forgot to mention that. I made that change quite a while ago, but yes it was intentional. I couldn't understand why subplots started at 1, since Python is 0-based. Combining it with subplot indices (plot[idx]), it made more sense to me to use 0-based axes numbers. Do you agree? If so, I could modify the docstring above

[jenshnielsen]

In general I agree that it's most logical for all arrays to be 0 indexed. I think the reasoning here is that pyplot.subplot(nml) is 1 indexed, as in pyplot.subplot(211) gives the first subplot of the 2,1 grid. That is an artefact of matplotlibs api being inspired by Matlab.

The biggest worry is that this will break backwards compatibility, @QCoDeS/core any opinion on this?

[nulinspiratie]

Ah right good point. In that case we could opt for the subplot kwarg to still be 1-based, while the indexing to be 0-based, similar to pyplot.add_subplot being 1-based, but the axes in fig, axes=pyplot.subplots() being an array, and hence 0-based. I hope this doesn't add more confusion though

[jenshnielsen]

Yes that is probably the best even if a bit confusing. It's consistent with matplotlib any way

[WilliamHPNielsen]

I would also opt for keeping consistency with matplotlib. That way I reckon we confuse the smallest amount of people to the least degree (although confusion is inescapable).

[nulinspiratie]

Alright just added a commit that changed this