Added USID_Writer

[saimani5]

Issue #56

[ssomnath]

@saimani5 - I made significant revisions to your functions in this commit. I haven't tested this function or the new changes though. I think it should work. One of the challenges was that USID >> sidpy.Dataset >> USID is not the best of tests since sidpy.Dataset allows a lot more diversity in the dimension types. Also note that sidpy.Dataset assumes that dimensions are arranged from slowest to fastest varying and this is not the default assumption in USID. This is the reason you will find the slow_to_fast being explicitly set to True now. Also, doing an np.array(object) effectively forces the data into memory and this would cause problems if the data is much larger than memory. This is the reason that sidpy.Dataset extends Dask.array rather than numpy arrays. Anyway, please do go over these changes. We should not / cannot close #56 until this function is heavily tested.