1-Moved the normweighmode to the beginning of the program.
In this way, the weight will be applied at the beginning
instead of just the final normalization step. Our tests
show that it improves the results when Max(JET, (PC+CV)2)
is used at the beginning. Most of the changes in this part
are in computePred.R file.
2-When the user provides a mutation list
as a file, it was giving an error. The error in apply(...)
part of normalizePredWithNbSeqsPCSelMult function. The bug
is corrected now and it produces results consistent with
the web server.

3-A small bug in gemme.py was also corrected.
If a mutation list file is given, the model is not 'simple'
and therefore, it will not produce a map of single point
mutations.

If the user does not provide a pdbfile and asks the program to calculate
structural-dynamical parameters like Bfactor, CV or DFI, the program will
give an error and quit. Before, there was not such a thing.

Reverted NA values from 2.0 to 0.0 bbc it was causing too much difference in the epistatic model maps.

If the query sequence contains a gap, createPDB() was blowing up.
This bug was corrected!

For now, alpha[i] is still 0.6 for the all residues to produce
similar outputs for all residues. In addition to this change,
corrected a bug related to the chain naming.