Misalignment between nalloc and num_poly for orthonormal polynomials

[goroda]

There may be an inconsistency in several routines (like axpy) regarding orthonormal polynomial expansions where in some cases num_poly may exceed nalloc. Need to fix this soon.