Facilitate running chunking on a single dataset in chunk.py

[eponvert]

Currently chunk.py expects a train / test setup. Allow user to only supply a single file, to chunk, output results and optionally evaluated output.