More robust sequence alignment based superposition

Jalview's existing superposition method constructs sets of pairwise mappings between locations in structures that map to sequence positions aligned in the MSA.

The existing method only proceeds to superpose structures onto the first 'reference' structure if there are at least 3 positions aligned in all structures, which is often not the case for multidomain proteins where only one or two representative structures are available for each domain across a family. Jalview should in this case select successive reference structures so that maximal matched pairs over the whole set of structures can be found.

e.g.
s1: AAABBB DDD
s2: ABBCCC
s3: CCC

with s1 as reference, s2 can be superimposed but not s3, but with s2 as reference, s3 can also be superposed onto s2 after it is superposed onto s1.