Calculate per document

Run T-scan on smaller parts e.g. single documents instead of the entire batch. After this is done these sub results should be merged. This should speed things up when updating a project (e.g. adding some new documents), make it easier to continue after a restart or crash (#73) and give more meaningful information about the progress (#55).

The issue names no files, tests, or entry points. Start by locating the batch T-scan processing flow and determine how individual document results can be stored and merged. Done means documents can be processed independently, resumed after interruption, and reported with meaningful progress while preserving combined results.

Written by the indexing model from the issue text.