Documentation fix: gh_sys_err

[maindlt]

Based on the Fortran code and its interface, GH_sys_err has number_GH entries. This PR updates the docs in this respect.

Before: number_GH: integer, the number of Gauss-Hermites

• GH_sys_err: a string of length 2 + number_GH floats, the systematic error applied to V, sigma, h3, ..., hN Now:
• number_GH: integer, the highest order of Gauss-Hermites
• GH_sys_err: a string of length number_GH floats, the systematic error applied to V, sigma, h3, ..., hN

Rationale:

1. in LegacyWeightSolver.create_fortran_input_nnls(), number_GH and the string GH_sys_err are written into nn.in, the latter with the expl. text [ systemic error of v, sigma, h3, h4... ].
2. In triaxnnls....f90's subroutine readobservations(), number_GH is read into hermax which is used to allocate kinerrscale(hermax), into which in turn GH_sys_err is read and applied to the moments. In triaxnnls_noCRcut.f90 this happens in lines 429-464.

[prashjet]

Sorry for hijacking this PR but it seemed like the best place to add this.

Tomorrow let's ask others to prrof-read, then merge both PRs and release :)