HMat: Provide new geometry, new distance and Gaussian covariance (+ associated matrix kernel); Resulting matrix has low rank off diagonal blocks (can be checked using testCompareClusteringRank).

HMat: ACA does not seem to work on 1/R matrix (TODO try different LRA techniques); Use SVD for now; Fixed default distance filename in testCompare...

# By Berenger Bramas (5) and Pierre Blanchard (2)
# Via Pierre Blanchard (2) and Berenger Bramas (1)
* 'master' of git+ssh://scm.gforge.inria.fr//gitroot/scalfmm/scalfmm:
  HMat: fix ChebSymKernel...
  HMat: Factorize ACA routines; Provided routine for recompression (ChebSym keeps its own recompression scheme, namely with storage in K); getRank can be found in namespace FSvd.
  test hmat clustering and rank
  use cell class in the antidiagonal at leaf level
  use inverse permutations
  new a plus equal
  add a custom well separated clustering

HMat: Factorize ACA routines; Provided routine for recompression (ChebSym keeps its own recompression scheme, namely with storage in K); getRank can be found in namespace FSvd.

HMat: Write distance matrix and kernel matrix in same test; Use distance matrix for clustering (in testCCL).