Add basic error checking for input csv samplesheet

[bkohrn]

EDIT by @maxulysse to add some checkbox to track the progress:
we should have proper check to see if data is correct for step:

• nextflow run . -profile test,docker --step mapping
  • other input data should fail
  • proper warning
• nextflow run . -profile test,docker --step markduplicates
  • fails
  • proper warning
• nextflow run . -profile test,docker --step prepare_recalibration
  • fails
  • proper warning
• nextflow run . -profile test,docker --step recalibrate
  • fails
  • proper warning
• nextflow run . -profile test,docker --step variant_calling
  • fails
  • proper warning
• nextflow run . -profile test,docker --step annotate
  • fails
  • proper warning

Some additional ideas from slack (edited by @FriederikeHanssen )

• If --step variant_calling or --step_annotation , then --tools should not be empty
• Bonus feature: check that the tools actual fit for the data (i.e. HAplotypecaller should cause a failure if only status tumor in samplesheet) #603

Is your feature request related to a problem? Please describe

It's frustrating to have the pipeline run through completely, only to tell me at the end that there are file name clashes based on me (as a new user of the pipeline) having set up my TSV file incorrectly.

Describe the solution you'd like

To help with this, it would be useful for the pipeline to do some basic error checking on the TSV file before starting any other steps; this means checking things such as: are values in columns where each row needs to be unique actually unique? Do the columns provided (at least) seem to match what should go into them based on the start point given (for example, does a file name look like a fastq / bam / bai file? Does the file actually exist?)? It would also be helpful if the pipeline supported a header row in the TSV file, as I suspect part of the reason I messed up was due to not having a header row to refer to.

Describe alternatives you've considered

Not sure if there would be any good alternatives to this; it's mostly a convenience feature to get the pipeline to throw an error before it runs all night and then needs to be restarted from scratch in the morning.

[FriederikeHanssen]

Sounds like a good idea to me. I suppose warnings could be added here:

sarek/workflows/sarek.nf

Line 468 in 085f1df

//TODO since it is mandatory: error/warning if not present?

similar to this

sarek/workflows/sarek.nf

Line 527 in 085f1df

log.warn "Missing or unknown field in csv file header"

.