Avoid get_option in innermost loop when determining quadrature points

This fixes a code generation performance bug in sum factorized code.
It also reveals that statistic profiling could be superhelpful to us.

python/dune/codegen/sumfact/tabulation.py

@@ -273,6 +273,7 @@ def quadrature_points_per_direction():
 
     # Quadrature points in per direction
     nb_qp = tuple(order // 2 + 1 for order in q)
+    _quad = nb_qp
 
     return nb_qp