Enzymes with 6bp recognition sites would cut every ~4kb, so a combination of two should work. Of course, the right combination depends upon your genome's GC content and di/trinucleotide frequencies. I don't know of a tool that does what you want, but your could use a kmer counter to determine 6mer frequencies, filter against a list of known RE sites (available from NEB or similar), select candidate REs based on the number of cut sites, and (optionally) perform in silico digestion to check the fragment size distribution.