fix(mpiarray): crash testing for an Ellipsis when given an ndarray index

[jrs65]

No description provided.

[tristpinsm]

These changes make sense, but as far as I can tell the new test and the changes to the ellipsis handling are not related. Should they be split into two commits?

[jrs65]

Yeah, they are related but it's a bit subtle. The problem with the old code was that if you pass in a numpy array as an array index it does an element wise comparison to Ellipsis and then complains that you need to use .all() or .any() on the result. So this test now checks that case and thus throws an error with the previous implementation.

[ljgray]

Any reason this hasn't been merged? Ran into the Ellipsis crash issue so having to use this branch to get around it.