Add support for 'collecting' csv/tsv outputs

[dfornika]

We produce several outputs on a per-sample basis that are simply a .csv file with a header and one line for the sample. Very often we want to collect those into a single .csv file that includes all samples in a particular analysis.

Add a --collect_outputs flag that will add collected outputs for any sample-specific csv or tsv files. Also support a --collected_outputs_prefix flag to set the filename prefix for the collected output files, with default value collected.