merge with kernels from MH's master thesis

TODOs:
  - Likwid marker in gather kernels should use the X_LIKWID interface.
  - Old kernels should set loop balance in KernelData structure.
  - AVX512 intrinsics for single precision must be implemented.
    - This affects all kernels using intrinsics
    - NOTE: gather kernels need special attention for gather intrincis.
  - Extend documentation by new kernels.