Cleanup `parsl` logging directory and make unique by default for concurrent usage

[d33bs]

@MikeLippincott brought up a challenge pertaining to parsl logging which typically occurs within the runinfo directory. We should clean up this directory and also make it unique per CytoTable run so as to enable concurrent usage. Without this, parallel processes may collide on the same logging directory and cause exceptions.

[MikeLippincott]

Hi @d33bs Thank you for opening this issue for additional context. I am having issues while running CytoTable on an HPC cluster where I am using multiple Nodes and cores to run CytoTable across multiple image-based profiles. When running CytoTable concurrently in the same directory, these directory clashes occur.

[d33bs]

Thanks @MikeLippincott ! While working towards a solution to this I found that Parsl currently hangs with a silent exception when customizing the run_dir path (new executors appear to reference previous executor files or status). I've opened up a Parsl issue to explore solutions on this with them. Parsl/parsl#3874

[MikeLippincott]

@d33bs has found a solution to this issue. I have tested on local and HPC and can confirm that it does indeed work:

convert(
            source_path=source_path,
            dest_path=dest_path,
            dest_datatype=dest_datatype,
            preset=preset,
            parsl_config=Config(
                executors=[HighThroughputExecutor()],
                run_dir=f"cytotable_runinfo_{uuid.uuid4().hex}",
            ),
            chunk_size=10000,
        )

Ensure that the parsl_config path gets passed into the args for the cytotable convert function. Ensure that the run_dir arg is passed through also. In our case, we pass in a dir that contains a uuid which makes an extremely low probability that parallel processes will write to the same dir.