when gaps included in x_single, there is an error in sampling

[sokrypton]

the ccmpred.parameter_handling.structured_to_linear() function linearizes x_single and x_pair.
by default it is assumed that x_single is a vector of size (L,20), if the flag nogapstate=False, it's assumed x_single is (L,21).

So far so good... but then if you try sample sequences using:
x = ccmpred.parameter_handling.structured_to_linear(...,nogapstate=False)
ccmpred.sampling.generate_mcmc_sample(x,...)

the "x" appears to be incorrectly parsed by function, assuming x_single is (L,20), and sampling is out of wack.

Appears this assumption (N_ALPHA - 1) is hard-coded in:
ccmpred/objfun/cd/cext/cd.h
#define X1_INDEX(i,a) (i) * (N_ALPHA - 1) + (a)
#define E1(i,a) x[X1_INDEX(i,a)]

then later in:
ccmpred/objfun/cd/cext/cd.c
void compute_conditional_probs(...) {
int nsingle = ncol * (N_ALPHA - 1);
for (a = 0; a < N_ALPHA - 1; a++) {cond_probs[a] = E1(i,a);}
...
}