Update mpl drawer backend to support unshared axes

[nkanazawa1989]

Summary

Experiment plotter can create a multi-sub canvas (axes) figure, but x and y plot range (xlim/ylim in matplotlib) must be shared among sub axes. This PR relaxes this limitation and allows a developer to write analysis class that can generate figure containing axes with different plot ranges.

Details and comments

Several figure options are updated to take a list of options for each axis. New options sharex and sharey are added to unlink the axis settings. These new options are True by default for backward compatibility.

[yaelbh]

I'm approving because my comments don't have to block approval. For each of the comments, you are free to accept or reject it.

1. Do you think "if there are multiple columns in the canvas" is clear enough?
2. Instead of the lists limit, unit, etc. it's better to have a single list where each item has data members limit unit etc.
3. Prefer zip over enumerate.
4. Is there a reason that scale is excluded, i.e., always the same for all the sub-axes?

[nkanazawa1989]

Thanks @yaelbh. I'll merge this PR.

Do you think "if there are multiple columns in the canvas" is clear enough?

I'm not sure if this is enough, but this is from the original documentation. I found a typo and fixed in 1217836. Thanks!

Instead of the lists limit, unit, etc. it's better to have a single list where each item has data members limit unit etc.

This is good idea. Probably we can introduce cleaner subplot mechanism in a follow up PR, i.e. introducing a dataclass that stores subplot configuration, and set multiple dataclass instances to write multi-axis figure. This PR aims at implementation of minimal functionality for Stark experiment.

Prefer zip over enumerate.

Yes, this makes sense if we have such configuration dataclass. Writing zip for every configuration is bit lengthy.

Is there a reason that scale is excluded, i.e., always the same for all the sub-axes?

I'm not sure if I understand. Current implementation doesn't support axis-wide scaling. This is because if we allow such setup we need to optimize the margin between axes to have tick labels otherwise we cannot stack axes with different scaling option. Implementation is not difficult but overkill for Stark experiment. I'll create a follow up PR based upon request.

[yaelbh]

Thanks @nkanazawa1989, can be merged when green.