flatten() always returns a new copy, but nearly all places that use
flatten() in the codebase don't need to pay that cost; the only place
that needed it did an additional copy. Fix them.
(To be honest it's not clear this will make a measurable difference, but heh.)

As a replacement, there's ravel(), which only makes a copy if the input
is not C-contiguous to start with, and reshape(..., -1) which never
makes a copy (as long as the input is already an ndarray), so prefer the
latter.

Note that in the changed snippet in mplot3d, the AttributeError would
never be raised, as x/y/z are converted to arrays and thus always have a
flatten() method.

PR Summary

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