Recovery bug in the case of spare rank failures

[Matthew-Whitlock]

The rank offset system used to determine which rank a given spare replaces currently assumes no failures in the spare ranks. If spare-rank 0 failed and user-rank 0 failed, spare-rank 1 should replace user-rank 0, but it will currently assume that this will be handled by spare-rank 0.

The line in question is here:
https://github.com/epizon-project/Fenix/blob/master/src/fenix_process_recovery.c#L515

This should be updated the check for the number of failed spare ranks of lower rank than my own, then subtract that from the calculated rank_offset. Probably needs to be updated in other parts of the function as well.