Multiple bug fixes

[radakb]

1. Fix buffer overflow error in fpocket with the -d option
   The buffer for env_atm filenames was too short for systems with > 9 pockets.
   There's probably a better way to fix this than just increasing the buffer size
   but this does the trick for all cases where I encountered the error.

2. Correct amino acid labels in fpocket descriptor output
   The order of amino acid names in fpout.c did not match the hardcoded order in
   aa.h. I spent two days looking for a reason as to why different amino acids
   would be found by fpocket and dpocket (see bug 3), but it turns out it was
   just a labeling error.

3. Correct the dpocket ASA calculation for implicit pockets near the ligand
   The implicit pocket descriptors were being computed with pdb_w_lig which
   resulted in different areas from fpocket because they are influenced by
   nearby ligand atoms. There was no issue when no ligand was present bc the
   pdb and pdb_w_lig data structures are identical in that case. I could not
   find any reason why ligand information would NEED to be present in that
   part of set_descriptors(). The parts where explicit pocket information are
   useful (example overlap calculations) should be unchanged.

   Note that a similar issue seems to persist for explicit pocket definitions
   and this might explain why the drug score for explicit pockets is always zero
   (a bug reported elsewhere).

I tested all of these changes by running fpocket and dpocket on the same complex and its synthetic "apo" pdb. I then diffed all descriptors in stdout from fpocket -d and in dpout_fpocketp.txt (only for the pocket with the largest overlap). The results were identical for atleast 10 random pdbs curated in the HiQBind data set.