Optimal secret re-assembly during recovery

[achamayou]

As per this discussion, the current secret re-assembly only attempts to use the first threshold shares passed to it (*).

An attempt may fail despite there being enough valid shares as input if there are duplicates, or bad shares in the first threshold values. We could implement the following two improvements:

1. eliminate duplicate shares
2. try permutations if the first attempt on a threshold free from duplicates does not work and we have enough additional values

[achamayou]

1. is of limited interest, members would need to access shares belonging to others and submit them too in order for this to happen.
2. is more interesting, because it would allow recovery even if a member submits an invalid share.