Change how assert_xydata() handles sets of data with different shape

[marty-larocque]

base.assert_xydata() uses np.testing.assert_array_max_ulp() to compare xy_data and xy_expected. When these two sets of data have the same shape but different data, assert_xydata() throws and AssertionError with a custom message.

However, when xy_data and xy_expected have different shapes, assert_array_max_ulp() throws a ValueError that is not caught. For our purposes, this is effectively a case in which xy_data and xy_expected do not match and should be handled the same way.

We should handle this case the same way and throw an AssertionError with a custom message.

[marty-larocque]

We decided to handle this case by throwing a ValueError, but still by throwing our own error with our own message. This is addressed in #219, and can be closed when that PR is merged.