GitLab

It lead to compilation errors on some systems.

 Now we can make use of the triangular equation to get limits on the lp loop

-Reduces number of gaunt calls
-Have to introduce llpmax to avoid wrong llp indices

eq.(28) of Phys. Rev. B 69 024415 (2004). Setting l_rdm=.false. restores
the old method, eq.(27). Both evaluated in real space. Leads to better
convergence if magnetization changes sign very often.

Rma olap

See merge request fleur/fleur!123