Fix an issue where compact beta sheet proteins had too few candidate protein atoms

[davidlmobley]

This fixes an issue where compact beta sheet proteins had very few candidate protein atoms.

The code which originally picked candidate protein atoms pruned the first and last three residues of each stable secondary structure element before selecting candidate protein atoms for further filtering. This makes sense for alpha helices, which are always at least 6 residues long, but beta sheets can be far shorter and still relatively stable, so this was overly aggressive for certain beta sheet proteins.

This fix adds a separate pruning value (two rather than three) for beta sheet proteins and fixes the failures we've encountered in selection of Boresch atoms in the protein.