To make the pipeline more robust, the global python variables for
various settings were removed where possible. Their values have been
moved to the configuration json file, and a jsonschema validation has
been added to the pipeline to make sure the configuration is valid.

The downsampling step using seqtk has been remove since it was not used.

The following additional changes were made:
 - Remove all --config values except CONFIG_JSON
 - Extend the config schema with the required and optional files that
   are supported
 - Add jsonschema validation of CONFIG_JSON
 - Remove global variables for scripts, add them to settings
   dictionary
 - Remove global variable for SAMPLES, use the settings dictionary
   instead
 - Remove support for multiple bed files
 - Remove support for multiple refFlat files
 - Remove support for downsampling of reads
 - Add json and jsonschema to the requirements
 - Update tests to work with the new config file

By default, GATK will group regions into 100 GQ bins,
with each bin containing only regions with the exact
specified GQ. This leads to large files wich are slow
to parse, while such a high resolution of GQ scores is
most likely not needed.

 - Remove pyfaidx from dependencies
 - Update readme
 - Fix some small spelling errors
 - Add tests for SCATTER_SIZE

Als include some tests for the content of the g.vcf file

Update the expected output of the vcf file in test_integration_run.yml, since
the P-value of one of the variants changed slightly with the new scattering.

dynamic is a special keyword in Snakemake that can be used to mark
outputs when the number of output files is not known before execution.
This is the case when using scatterregions, since the number of outputs
depends on the size of the reference genome and the size of the
scattered chunks.

See
https://snakemake.readthedocs.io/en/stable/snakefiles/rules.html#dynamic-files
for details

This fixes a rare bug where picard runs out of disk space on when spilling
reads to disk, because it uses the internal java temp dir as a location. This
closes issue #42

Deprecate conda, make pipeline more general

See merge request !13

# Conflicts:
#   Snakefile

When a repository changes its Suite value from 'testing' to 'stable',
this change needs to be accepted or the repository will not be used.
This leads to errors in the gitlab-ci testing since the latest version
of pip3 is not installed.

For some reason, pip fails in the python3.6 container on install
python>=3.5. This makes all gitlab-ci tests fail.

Fix typo in resource requirement.

See merge request !12

Hutspot sometimes fails on the fastqc analysis due to Java heap space
memory errors. It is not possible to set the memory usage by fastqc
directly, since this is done by the perl wrapper script. However, the
amount of memory that is set by the wrapper scales with the number of
threads that fastqc uses. Therefore, the number of threads was
increased, effectively increasing the memory fastqc can use.