Use pre-compiled code for CellwiseInverseMassMatrix

[kronbichler]

By looking at a profiler output I realized that CellwiseInverseMassMatrix would not use the optimized pre-compiled version of the code with compile-time loop bounds, but rather the slower path with run-time loop bounds. It turned out that we forgot to add the respective call to the templated/precompiled code in CellwiseInverseMassMatrix. While there, I also simplified the calls slightly as the fe degree can be read in a more central place.

[peterrum]

The following tests FAILED:
	2545 - matrix_free/inverse_mass_02.debug (Failed)
	2580 - matrix_free/advect_1d.debug (Failed)
	2581 - matrix_free/inverse_mass_01.debug (Failed)
	2584 - matrix_free/advect_1d_vectorization_mask.debug (Failed)
	2640 - matrix_free/inverse_mass_03.debug (Failed)
	2689 - matrix_free/advect_1d_system.debug (Failed)
	2691 - matrix_free/inverse_mass_05.debug (Failed)

[kronbichler]

Thanks, I had only executed the relevant application but not the other code path. Good to have some tests. 😄