split kernel computation into two parts

[Dengda98]

At present, grt_(static_)kernel function consist GRT computation and QWV assembly, which is not convenient for Surface Wave mode calling.

This two steps can be devided into two functions.

Also, GRT and QWV can be GRT_MODEL1D members ?

[Dengda98]

GRT have been moved into struct GRT_MODEL1D in #146

But for the QWV and related sum_J arrays, I think need another struct to specifically manage the wavenumber integration process. As comparison, now GRT_MODEL1D only consist physical quantities and Generalized R/T matrix.

[Dengda98]

kernel function splitted in #156 , while the QWV will be considered later.