Masking coordinates flattens 2D data

Currently when masking coordinates the arrays are flattened so that 'inner' masked points can be removed, e.g.:

[[1.0, 2.0, 3.0],
 [4.0, NaN, 6.0],
 [7.0, 8.0, 9.0]]

- > [1.0, 2.0, 3.0, 4.0, 6.0, 7.0, 8.0, 9.0]

However in some cases the data isn't really two dimensional, they are orthogonal coordinates, so that we are more likely to have:

[[1.0, 2.0, 3.0],
 [NaN, NaN, NaN],
 [6.0, 7.0, 8.0]]

and we want to keep the 2D structure:

[[1.0, 2.0, 3.0],
 [6.0, 7.0, 8.0]]

In order to fix this we need to be able to discern between the two cases. Ideally the second case wouldn't actually even be a 2D structure to start with, but this requires the data model changes discussed in JASCIS-248.