Fix Matplotlib renderer.M deprecation

[splch]

Summary

When plotting a bloch sphere using Matplotlib version 3.4.2, the following deprecation warning is given:

/opt/conda/lib/python3.8/site-packages/qiskit/visualization/bloch.py:69: MatplotlibDeprecationWarning: 
The M attribute was deprecated in Matplotlib 3.4 and will be removed two minor releases later. Use self.axes.M instead.
  x_s, y_s, _ = proj3d.proj_transform(xs3d, ys3d, zs3d, renderer.M)

By changing x_s, y_s, _ = proj3d.proj_transform(xs3d, ys3d, zs3d, renderer.M) to x_s, y_s, _ = proj3d.proj_transform(xs3d, ys3d, zs3d, self.axes.M), the warning is fixed.

Details and comments

Version information:

Qiskit Software	Version
Qiskit	0.27.0
Terra	0.17.4
Aer	0.8.2
Ignis	0.6.0
Aqua	0.9.2
Matplotlib	3.4.2

[jakelishman]

We made the same change in qutip/qutip#1556. Given that this is the same file (give or take), there's another change about deprecated automatic adding of axes to figures (see qutip/qutip#1503) that might be necessary as well?

[mtreinish]

I looked at this problem back when matplotlib 3.4 was first released here: #6087 the issue we hit is that there doesn't seem to be a way to thread the needle on our current minimum supported matplotlib version and fixing the deprecation warnings. The only way we can fix this is deciding to bump our minimum matplotlib version

[jakelishman]

In QuTiP we parsed the matplotlib version from matplotlib.__version__, and then just branched on the version:
https://github.com/qutip/qutip/blob/833cea2be5a0ac503e4fc925a0a0313718f76e85/qutip/bloch.py#L23-L29

It's not the most elegant, but I was of the same opinion as you about maintaining a good period of compatibility. For reference, mpl 2.1 is January 2018, and 3.0 is September 2018. If 0.19 is planned for late (late late) September, then bumping to mpl 3.0 will keep a 3-year compatibility window, and ease a couple of the branching requirements.

[mtreinish]

Since this was opened we've had to make a bunch of changes to the bloch sphere visualization to fix compatibility with newer matplotlibs (v3.5.0 kind of forced our hands because it broke compatibility with other things). This was incorporated into the those changes which were released as part of the 0.19.0 release. So, this PR no longer applies to the repo, so I'm going to close this now. Thanks for tackling this earlier even if didn't end up merging in this form and please feel free to re-open this if I'm missing something or there is more to discuss on this.