Allow numpy arrays in markevery

[ImportanceOfBeingErnest]

PR Summary

This line

matplotlib/lib/matplotlib/lines.py

Line 583 in 9a9c8b8

if self._markevery != every:

prevents numpy arrays to be used as input to markevery, Because The truth value of an array with more than one element is ambiguous.
That's actually a regression that came in via #4091.

I think we would like to support numpy arrays wherever possible. So this PR just removes the check and sets the artist stale on every call to set_markevery.

Also creates tests for all possible cases that markevery can take.

PR Checklist

• Has Pytest style unit tests
• Code is Flake 8 compliant
• New features are documented, with examples if plot related
• Documentation is sphinx and numpydoc compliant
• Added an entry to doc/users/next_whats_new/ if major new feature (follow instructions in README.rst there)
• Documented in doc/api/api_changes.rst if API changed in a backward-incompatible way

[timhoffm]

test_lines/test_markevery[png]

got:

expected:

diff:

[ImportanceOfBeingErnest]

Wow, this was hard - because the tests ran perfectly fine locally. Turns out I stepped directly into numpy/numpy#16023 (np.array(["0"]).astype(bool) is False in all but the very recent numpy 1.18.3).

[timhoffm]

I think this would be more readable:

    cases = [
        (None, "11111111111111"),
        (4, "10001000100010"),
        ...
    ]

    for ax_test, ax_ref, (markevery, pattern) in zip(axs_test, axs_ref, cases):
        ax_test.plot(t, y, "-gD", markevery=markevery)
        ax_ref.plot(t, y, "-gD",
                    markevery=np.array(list(pattern)).astype(int).astype(bool))

[timhoffm]

Optional: There is no gain from random numbers here. You could just use y = t**2 or any other simple function.

[efiring]

@timhoffm's suggested test changes look easy and worthwhile so I'm holding off on a review, but otherwise this looks good.

[tacaswell]

I agree with @timhoffm that grouping the cases the other way would be clearer, but I am not going to hold up merging over it.

[efiring]

Let's get this in now. @timhoffm, if you would like to put in a simple subsequent PR to improve the tests, I'm sure that can also go in quickly.