Remove separation of PE/SE for deduplication

[jfy133]

After some reports - it seems mixing PE/SE is more common than I thought.

The separation is actually highly specific for runnign with DeDup and is not necessary for markduplicates (which is also more well known and used).

DeDup can actually accept BAMs with it mixed, but will not necessarily work 'as expected' - so better to allow and throw a warning that it is not optimal.

[jfy133]

Furthermore, this means this can be merged during lane merge.

However will need to make sure it handles paired-end mapping correctly.

[jfy133]

@aidaanva and I had a quick discussion and as it's the most practical coding-wise, but also in context wise, we will always merge PE/SE of same library after AdapterRemoval and prior mapping. The important thing here is that we will then force a 'SE' SeqType to this data from then on.

This means that even though you do have PE in your library, you cannot then use the bwa sampe.

The alternative is to add an additional merging step after mapping to merge a both sampe and samse library, however this is a lot of code and importantly, for aDNA, we almost always have such short data we will always merge PE reads anyway. We rarely want spatial information from mate-pairs, and if someone did, it is even less likely they would also have sequenced SE data.

We can revisit this in the future if becomes a request.