Add an explicit sort function

[dougbrn]

Certain user workflows will require their data sorted in a certain manner. The most obvious being their data sorted by time, per lightcurve (So a first-order sort by object_id, and second-order by timestamp). We should provide a sort function that gives them the ability to do this. There is some risk that this is resolved automatically by the refactor, but the need to keep lightcurves in object_id groups with each sort suggests this might need to be a function beyond the scope of a dd.sort_values call

[dougbrn]

https://docs.dask.org/en/latest/generated/dask.dataframe.DataFrame.sort_values.html

sort_values is able to sort by multiple columns. I'm not 100% sure if it's happy with one column being the index, as will be our use case most of the time. But there's a strong possibility inheriting sort_values via the refactor (#191) will resolve this automatically.

[dougbrn]

We have sort_values now, but there may still be an argument for a sort function that always sorts first on index (to preserve divisions), but then sorts by a second-order key within each index grouping. You can do this with sort_values, but asking users to use it for these use-cases may often result in the user just blowing up divisions by doing something silly. Perhaps even overriding sort_values for the source table is the right move here.