Remove redundant coords assignment; test for dims and coords

[milancurcic]

I don't know why this fixes this index dim on some systems, but it seems like it does. I added tests to ensure that only obs and traj are in dims and coords.

@selipot @philippemiron please check on your ends.

Closes #323.

[selipot]

Ok index is gone.

But why not having like for the GDP the coordinates time(obs) and ids(traj) instead of obs(obs) and traj(traj)? I find this not really comprehensible. The coordinate you have currently called obs is effectively a time variable and the traj is a unique identification string.

[philippemiron]

Works for me too. Only comment: should we put a message to notify the user when the file already exists in ~/.clouddrift/data/?

[milancurcic]

Yes, let's discuss the naming convention.

My preference may be time(time) and id(id). Unclear to me why name coords differently from dims, as they map 1-to-1 in our case.

[philippemiron]

And I agree with Shane. I'm not sure I understand the addition of obs and traj as coordinates either.

[milancurcic]

As discussed on the call, we will rename obs(obs) and traj(traj) coordinates to time(obs) and id(traj) respectively. Dimension names remain the same.