Feature/duplex

[Adoni5]

Two implementations of simple duplex targeting

• Initial implementation of the "chill" accept if the previous read was deliberately sequenced, and this one otherwise wouldn't be, which has been included to account for sequencing duplex with no_map as the decision.

• The slightly more complex previous alignment was sequenced, and this read aligns to the opposite strand on the same contig.

Both of these modes break the chain of stop_receiving if the previous read was only sequenced if it was potentially duplex, by checking the new duplex_override decision, on the duplex_tracker.

Other things of note:

1. Version is included in the printed output at the top of the logs
2. Small change to an error in mappy.py index extension checking, which not includes the extension that is incorrect for clearer error messaging
3. Added a new decision of when we override a read at the start of a readfish run, if the translocated portion is of unknown length

[mattloose]

It has been tested live.... waiting to interpret those tests.

[Adoni5]

Checked the simple duplex

Of 220,000 Duplex reads after 4 days of running - we sent unblocks to 42,886 to the one or both of these reads parents, (29051 reads had an unblock sent to one of the parents, 13835 to both of the parents).

However looking at this by read lengths it's probably fine -

And over the whole width of read lengths:

Note - 10 base bin width for both plots.

It's roughly 0.7206% of total duplex bases, so I think this is working!

To avoid this languishing, I'm suggesting we merge this in and mark it as highly experimental.

[Adoni5]

This works - @alexomics and @mattloose I'd like to merge so we can address #347 and avoid this languishing so long it no longer works

[mattloose]

Yes - we know that this works now. I'm happy to merge it.

[Adoni5]

@alexomics blocked by your requested change for being certain about Strand which has been addressed!