Famli

**Aspera**

By using Aspera, we completely avoid having to write to the root partition, which is limited to 8G in AWS Batch. This was previously causing a fair amount of failures which should now be eliminated.

**Minimum read length**
A minimum read length was added to the fastq quality trimming step. The default value selected was 30bp, which corresponds to 10aa. With a default minimum alignment score of 20, there should be no case in which a trimmed read with < 30bp had ever produced an alignment in previous versions of the code. Therefore the output should be the same as the previous release, as long as the minimum alignment score was >=10.

Fetching data from SRA via `fastq-dump` writes files to `/root/ncbi/public/sra/`, which cannot be changed readily and which directly impinges on the space available in the partition used for Docker. This release includes a simple fix to delete those temporary files in between each file that is fetched. This mostly helps for very large samples that consist of multiple combined files.

There is also a bugfix that gracefully handles the case when 0 reads are aligned against the database.

Refactored the alignment batch processing to limit the number of alignments that are read into memory at any one time. The previous release was running into memory errors at ~220M alignments on a machine with 72G of RAM available, and so I hope to overcome that limitation with these improvements.

Added a parameter (`--batchsize`) that optionally allows the user to process alignments in batches of a customizable size. When enabled, this feature will perform filtering and read reassignment on subsets of the data, combining the results for the final set of results.

This enhancement was initiated because of memory errors when processing ~50M reads at a time.

There is no change in functionality when the `--batchsize` parameter is not set.

Instead of calling `famli.py`, now call either `famli.py align` or `famli.py filter`. The former command provides the same functionality as the former `famli.py`.

Also mask any non-ATCG characters as 'N' in the FASTQ (which fixes a breaking bug in `fastq_quality_trimmer`)