Batching of samples for create import TSVs

[kcibul]

Processing an exome takes ~1 minute, which means most of the time is spent on spinning up a VM, pulling docker images, etc. This is not very cost efficient. This PR allows for a batch_size to be set and then each task processes that many samples as a unit. The default is 1 which yields the current behavior, but in exomes I have set it to 20 and seen the cost to ingest drop dramatically

The GitHub PR makes it look like a lot has changed but really the changes are:

• a new parameter
• a new task to turn the Array[File] for the VCFs into set of FOFNs (file-of-file-names) similar to how we split up intervals
• a loop in the actual Create TSV task to loop over the files in the FOFNs. For SA mode we copy down each file, and for non-SA mode we rely on the fact that localization is optional and we read them directly anywy

[ahaessly]

👍