Use worker run IDs instead of worker versions

Followup of teklia/arkindex/backend!1906

Several import tasks will be broken as publishing results with worker version is not supported anymore by arkindex.

We need to convert the usage of WORKER_VERSION_ID by ARKINDEX_WORKER_RUN_ID and use it as worker_run_id on all ml results publication endpoints.

This affects PDF text import (datafile & S3), transkribus import (through pagexml publication).

Mentions of worker version in the git import should not be affected as this is related to the pubication of new workers !