Decouple `OffTargetDetector` from the `prymer` data structures

[msto]

It'd be helpful if OffTargetDetector did not require Oligo or PrimerPair objects as input.

The off-target detection does not use any of the primer-related metrics that appear on these classes (e.g. tm, penalty), and this design constraint makes it challenging to reuse this functionality in other contexts.

I think the detection methods could be re-written to accept strings (or string tuples, for pair checking) as input. We could also consider accepting a Protocol for any class that includes a sequence or bases attribute.

[tfenne]

My 2c is that this is likely an "and", not an "instead of". I would object to there being two sets of functions:

• One that takes primers/primer pairs
• One that takes just sequences (or pairs of sequences) as strings

However, long term I think it is useful to have functions that take the richer objects. There are features I'd like to implement in OffTargetDetector that would require them (or a much more complicated Protocol), including:

• Filtering out the on target locations of primers when returning "off target" mappings
• Being able to filter off-target mappings by thermodynamic properties (e.g. the off-target mapping's Tm is at least 10C lower than the on-target Tm)

The functions that take just strings would not be able to do this, but that doesn't mean we shouldn't have them as an alternative.