Group samples: split summary stats

[ewels]

Branching off from #542 as a similar yet distinct issue.

In some cases we have repeated outputs from a tool from multiple steps in an analysis pipeline. The classic example is to have results from FastQC before and after read trimming.

The current solution for this is to run the module twice, using path_filters (see example in docs). This treats the two sets of outputs as if they were from totally different tools. It works relatively well, but it's not super easy to use and has some limitations.

Proposal in this issue is to have a new config feature called sample_split_groups that works very similarly to #542 / #576.

Instead of merging summary statistics in tables as in the above feature, stats should be shown in separate columns. This behaviour will be identical to the current practice of running modules multiple times.

For plots, all data will be shown in a single module section (unlike running modules multiple times). Data will be shown as multiple datasets, which can be switched between. This is different to merging groups, as it'll be switch- behaviour and not toggle- behaviour.