Set process name for long-running processes

For long-running processes it can be convenient to have a short and
human-readable process name. For example, in our typical server
configuration we have quite a few Mutalyzer processes running with very
long process names, making ps and top output hard to parse.

It is assumed that in production environments the webservices and
website are ran in a separate WSGI server such as Gunicorn which can be
configured to set the process name.

The setproctitle dependency for this feature is a C extension and we
therefore make it a soft dependency. Construction copied from Gunicorn:

https://github.com/benoitc/gunicorn/blob/5bc13be79ecaf8f1027764c600023f674ebcf5b2/gunicorn/util.py#L55

Don't report ext*? when variant RNA has stop codon

With the change introduced by #65 we forgot if the variant RNA has an
alternative downstream stop codon and therefore always reported ext*?
when the original stop codon was removed.

Fixes #145

Release 2.0.15

Degrade gracefully on HTTP error from Entrez

Add gene feature to genbank file without version

With this change the genbank parser no longer discards incomplete genes
directly but keeps them as long as they have complete features
annotated.

For example, the PIK3R2 gene is annotated on NC_000019.9 (or a slice) as
4973..>22328 with two RNA entries. One of these, however, is complete so
it would be a shame to discard the entire gene.

This genbank file is incomplete and incorrect anyway, but this was
not the mistake we want to test.

Suppress warning about env variable in tests

Create legend only after gene model enrichment

This fixes a bug where transcripts created from CDS by construction did
not show up in the legend because the legend was created before that
construction.

Fix docstring in sliceChromosome API method

Configuration by environment variable is optional

Parse genbank file without VERSION field

Partial fix for https://humgenprojects.lumc.nl/trac/mutalyzer/ticket/188

Speedup NCBI mapview file import

Instead of querying the existing mappings for overlap and either
updating or inserting depending on the result, we now delete
overlapping mappings first and then only insert.

This is roughly twice as fast. But of course still a horrible
setup compared to some kind of UPSERT functionality which is
unfortunately missing in current PostgreSQL.

Drop BatchJob.download_url column

This is now created on use, by #111.

Fixes #112

Drop TranscriptProteinLink database table

This data is now in Redis, by #94.

Fixes #95

Update requirements

Readable name for batch notification sender

Release 2.0.14