ATTENTION:
use --rrecover and --wrecover options with catution:
first: run treekin with -w, i.e. let it diagonalize the input matrix once and write
       the eigenvalues and eigenvectors to the corresponding *.{evals,evecs}.bin files
second: run treekin with the -r option, i.e let it read those *.{evals,evecs}.bin files
        and do ONLY the iteration (no diagonalization)
NOTE: when using --rrecover and --wrecover option in combination with the -a option,
      treekin assumes that the absorbing state between consecutive calls of the program
      are the _same_ (since the information about the absorbing state influences the
      transition matrix).
      If you change the -a option while calling treekin with --rrecover, treekin will
      produce junk output (since the transition matrix that has been diagonalized earlier
      is different from the current one, due to the different absorbing state)