Gracefully handle when no reads map to a chromosome

[kelly-sovacool]

While running a test dataset with a subset of reads from only chromosome 22 (see #27), rule varscan_single failed for all other chromosomes because no reads mapped to them so there were no variants to call. (The varscan command within that rule didn't error, but bcftools errored right after due to the empty file.)

We could prevent this further upstream after rule split_bam_by_chrom. If this rule produces a bam file with no reads for a given chromosome, remove it from the chroms list so downstream rules don't use it. Will probably need a checkpoint to accomplish this because the DAG will have to be reevaluated with the new chroms list after all split_bam_by_chrom instances complete.

Temporary fix I'm using now is to manually edit the config file to include only chr22 and then manually run kickoff.sh so it doesn't override my custom config (see #29).