Cache reference downloads / index builds, provide snakemake profile with default cache dir

[tedil]

In order to make the snappy-pipeline more flexible and portable, when #512 is done, the reference may be downloaded or linked automatically, dependent indices can also be built or linked automatically, and the results can be cached in the SNAKEMAKE_OUTPUT_CACHE to be made available between workflows.

[tedil]

This also helps with reproducibility, because the way to obtain such data is included in the workflow.