Improvements on final MultiQC reports

[cnluzon]

Some improvements on the final MultiQC reports:

1. Allow for zero reads at some steps: barcodes that didn't work will still make it to the final report, even upon failure of the run. I think it is OK if the pipeline crashes on those, as there is something unsuccessful - final bigWigs cannot be generated for an empty barcode. However, it is useful that this does not prevent the final report to be produced, and that these zero values are explicitly shown in the final report - A line with zeros and NA values will appear. If no_bigwigs rule is triggered, the run will still be successful. For this, I forced the Picard step to produce a dummy file, and handle some zero exceptions on the stat summarization steps.
2. Make all unknown estimators show up as NA to make it consistent. Je puts a question mark on empty library sizes, library size estimation returned None, so now it is all NA for everything.
3. Carry barcode sequence all the way to the stats_summary.txt file. I don't think it needs to go on the HTML report output maybe as there are a lot of columns already, but it is still nice to have it in the final table - sometimes we want to dig deeper on barcode performance statistics.
4. Some styling changes to the barplots: smaller base font size, better colorscale and also cleaning the labels (replacing dashes and underscores with spaces on the final plot) and wrapping them in several lines if they are too long, so the final figure is not messed up if someone used very long names for fastq/conditions.