Merge branch 'fix/smooth-kernel' into fix/clean-up

* fix/smooth-kernel: (25 commits)
  We now use the fmm operator to calculate the interaction lists
  Add a Gaussian kernel with the option of changing the kernel coefficients
  Add two defines   1) LOW_RANK_USE_SVD to enable truncated svd rather than partial aca   2) LOW_RANK_CHECK to check the reconstruction error
  problem with fft and settings is fixed
  Add a modified 'one_over_r' matrix kernel that adds a diagonal value.
  Add 'smooth_tag' for smooth kernels + trait to detect at compile-time if a kernel is smooth.
  Update smooth kernel implementation:
  Update smooth kernel implementation:
  matrix_kernel: complete gaussian matrix kernel to add an 'evaluate' member function
  Modification of the default coefficient of the Gaussian kernel. Test how to modify the coefficient in the main code.
  added comments
  Same as previous commit
  added functions to print the settings of the fmm operator (approximation, optimisation, kernel)
  Added Gaussian kernel
  We have a loss of precision with the low-rank approximation in uniform and Chebyshev compared to dense and fft
  add symmetric tag
  Correct format for doxygen documentation
  Code works if Gaussian coeff is 1.0 otherwise the results are wrong if we set
  The direct calculation is validated (verified from pyhton)
  delete dead code
  ...