parallelize API interaction once #6 is done

until the import

importing a large tree (e.g. 500-1200 categories) takes a fairly long time currently because everything is completely sequential. It's not possible in a different way because the sync does not "understand" the tree ordering to do stuff in the right order.
But once #6 is done it could be possible to parallelize API interaction (e.g. in batches per tree depth).

not critical as massively huge category trees aren't the norm.

Start by reviewing issue #6 and the current sequential category-tree import flow. Determine how the API interaction can be grouped by tree depth without breaking ordering. Done means large category trees can use parallel API batches while preserving the required parent-before-child sequence.

Written by the indexing model from the issue text.