fix: general fixes and optimizations

[FelixMoelder]

Since updating all integrated wrappers the workflow fails at certain points. Besides we write a lot of intermediate fastq and bam files without deleting them. For large samples this might end up in over a hundred gigabytes of additional disk storage. I therefore marked all of these intermediate results as temporary only keeping merged fastqs and recalibrated bam files.

List of changes:

• updated datavzrd to latest version
• updated vep annotate wrapper as it still used vep v109 while vep cache already already used v110
• set memory for picard to 3gb
  • the faq recommand to use 1GB per 250.000 reads in memory with length of 100bp. Per default there are 500.000reads in memory and I assumed a read length of 150bp ending up with 3GB. In addition broad uses internally at least 2GB (https://broadinstitute.github.io/picard/faq.html)
• downgrade gatk baserecalibrator wrapper as baserecalibration fails.
  • the error seems to be caused by gatk as the error says that the known variants gtf file can not found which is previously copied to a temporary directory by the tool it self. Considering the documentation there has nothing changed how gatk is run so this seems to be a bug.
• updated ensembl-variation wrapper as the previous version failed due to downloaded index files
• updated reference to build 110 to be in line with vep versions

Wrappers versions are not final, yet. They have been updated but we still need to wait for a new snakemake-wrappers release.

[FelixMoelder]

This is ready to review. Still, before merging we should update datavzrd once available as the was a rendering error fixed.