PR Summary

Previously, if an animation-generating subprocess (e.g. ffmpeg)
failed while data was piped in, we'd get a ValueError because we'd
communicate() with the subprocess once in grab_frame(), to generate a
first error, and again at cleanup time, in the finally: clause, but that
second time, the subprocess' stdout and stderr had already been closed.

Instead, don't do anything in grab_frame() and let the finally: clause
handle the errors.

Also replace the nondescript RuntimeError by a CalledProcessError, which
is more descriptive, generates a proper error message by itself
("Command 'foo' returned non-zero exit status 42") and is more easily
introspectable (it includes stdout, stderr as attributes). Given that
we would previously sometimes throw a ValueError instead of the intended
RuntimeError, I think the API change is worth it.

Un-xfail a test.

Builds on top of #12368. Closes the second half of #9205.

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