US-4.4not implemented Optional incremental pipeline runs
As a user who runs the pipeline monthly and would like the option to only re-process new bank statements I want the pipeline to support a `--only-compute-new-tnx` flag that skips already-processed CSVs so that the pipeline runs faster when I know nothing has changed.
Acceptance Criteria
- Without `--only-compute-new-tnx`, the pipeline recomputes all journals from scratch every run (existing behaviour, unchanged).
- With `--only-compute-new-tnx`, the pipeline tracks processed file hashes and skips CSVs whose hash has not changed.
- The journal output is identical whether run incrementally or from scratch.
- If a hash mismatch is detected (file changed), the pipeline re-processes that file and warns the user.
No DAG diagram: Not yet implemented — optimisation flag; the data path is the same as US-4.1 but with caching/skip logic around unchanged files.
Demo video coming soon