Read mda data into memmap

[jchutrue]

My files are fairly large so I usually work with numpy memmap's to deal with the out-of-core-memory problem. It would be nice to have an option allowing readmda() to return a memmap instead of loading the data into main memory.

[tjd2002]

Great idea. The structure on disk is really simple, so this should be pretty easy. PR’s accepted!

[jchutrue]

Great! By the way, was there a reason we use column-major order instead of row-major? Numpy is row major by default so just curious.