Add checkpointing

[aaronmussig]

If an output file exists and the hash is the same then it should be skipped.

Notably, a few steps will need additional files created i.e. FastANI. In cases like this where multiple threads are processing data then the writer thread should write these results to disk (e.g. every 500 results).

Perhaps each major/minor step could interface through JSON/YAML files which would serve nicely as a method for checkpointing.