This is an attempt to generate more interesting output when the genome
has extra transgenes.

Hopefully this will be more efficient on pisa.

Pisa has more memory. Trim and dedup seems to be a bottleneck.

See https://github.com/deeptools/pyBigWig/issues/97

Interface seems to have changed. There may still be missing changes.

Made genome configurable for GRO-seq, tried to debug rpy2 uncachable
exceptions.

Could the `--quote` option have been another solution?

Also changed the way to cleanup temporary dir when making bigwig files.

We added all_si_22G and all_si_26G, who have a signature (more
informative than all_si) but are not dependent on being antisense to
some annotations (avoids being lost when annotations are apparently
incorrect, like some histone genes 3'UTRs ending too early).

Also removed indexing of annotation files. For some unknown reason,
tabix started complaining about wrong sorting, and this is not used any
more anyway.

22G should actually be 21-23 nt long.

The 22G are actually too few. We will try to replace this with 21-23G.

This is not yet fully tested, and may have effects on other pipelines.

Also cleaned and commented code.

This is to limit clutter from old analyses.

Snakemake 5.2.2 now checks that log file names have enough wildcards.

Plotting 5UTR, CDS and 3UTR on the same graph can help discover
interesting behaviours.