Further optimizations for swifttcr

[rhinowarrior]

Possible to write the output of apply_results.py and merge_pdbs.py to memory and only create the top models that are in the clustering.txt.
Right now the results of apply_resutls.py and merge_pdbs.py are written to files which takes a lot of time and slows down the program