Switch {PERMUTE,TRANSPOSE_SIMPLIFY}_BFS compression steps to Rust implementation

The Java implementation of PERMUTE_BFS does not fit in 2TB of RAM anymore

This requires changing PERMUTE_LLP to work from the base graph + composed order (instead of BFS-permuted graph + LLP order) because the Rust implementation does not write the BFS-permuted graph on the way to the BFS-permuted+transposed+simplified graph