Add selobs/seltrajectory method

[gauteh]

Often you want to iterate over trajectories and get the data in the time vs variable format. As long as you only have one trajectory you can reduce the observation layout to just 1d time layout. Would be nice if this works with groupby as well.

ds.traj.iseltraj(0)

gives a ds in (trajectory, time) format (except trajectory is size 1).

ds.traj.grouptraj().map(lambda t: t....)

[gauteh]

Will be fixed by #104 and the to_1d() method:

ds.isel(trajectory=0).traj.to_1d()