Cmake fix

[mtaillefumier]

• The SIMD tests are now compiled and executed (issue DetectSimdAndAlignment detects AVX512 on AMD EPYC (znver2) #684 )
• Naming convention for the cmake config files is uniform (issue inconsistent naming in CMakeLists #691)
• Updated the commit version of lemon (PR update lemon commit when installation via FetchContent is fixed #690)
• Solve the installation directories inconsistencies (issue inconsistent naming in CMakeLists #691)

[mtaillefumier]

@chaoos : I would appreciate your feedback on this PR. It needs to be merged soonish as it fixes few issues with include directories.

[kostrzewa]

@kottnad could you please review this and merge?

[chaoos]

@mtaillefumier Looks good to me.

Only small concern I have is that you changed multiple check_cxx_source_compiles into check_source_runs. That might lead to problems when not building on a target node. This is no problem on daint, but maybe on another cluster. Is there a possibility to bypass the SIMD detection and provide values explicitely?

[mtaillefumier]

True. I just do not know how to check which instruction is effectively supported during compile time without running it after compilation. check_cxx_source_compiles only checks that the code compiles not that the instruction is supported on the machine at runtime, that's why my initial implementation was defective in practice.

The best solution would be to check if the instruction set is supported at compilation time without running the executable but I am not sure if it is possible in practice.

Is there any case of cross-compilation ?

[chaoos]

I understand, please go on. We can think about adding a bypass when the problem arises. For me this can be merged, thanks for the changes! 👍