Modify file carton to deal with duplicates

[albireox]

This is a major refactor of the FileCarton class returned by the get_file_carton() function, mainly to deal with duplicate rows for an input identifier. This can happen when a single identifier (e.g., a Gaia DR2 source_id) is associated with more than one catalogid.

When this happens the new code will rank the duplicates by distance (assigning distance=0 if the distance is null, as for phase 1) and select the one with the lowest distance. Internally this is done via a subquery with a window function that partitions on target ID.

A corner case is when all the duplicates have been added as part of phase 1 and all the entries have null distance. In this case the code will select the entry with the lowest catalogid.

Apart from some code cleaning, the new code also changes a few things:

• The file data is loaded into a temporary table without using the copy_data() function. Since the file table can be dumped into a CSV safely, it does that internally and copies the CSV, which seems to be a bit faster for large tables.
• The union of all the subqueries for the different parent catalogues is run as a CTE and then the results are distinct'd on catalogid. This prevents duplicates when the same catalogid is returned by different subqueries.
• Disables sequential scanning for the file carton queries.