General Hidden Markov Model Library Code

stores the length of the viterbi path ending in each state

much easier backtrace and allocation for HMMs with silent states
a little bit slower but saving 100 lines of ugly code

avoid comparing doubles in ghmm_dmodel_viterbi if comparing int is also possible

exhange phi pointers instead copying the values

return the path length in ghmm_dmodel_viterbi()

helps parsing of the result in python for HMMs with silent states

remove one python loop in getGeneratinStates

remove trailing whitespace

don't loop in python, when parsing the viterbi path

integer to double conversion is mostly without loss

implements python list <-> double array conversions
move array based helpers to wrapper_arrays.i

implement list to int array and vice versa conversions in the ghmmwrapper
15 fold speed increase for large lists/arrays