Slow for large instance compared to alternative package

Thank you for coding up the algorithm! I noticed it is slow compared to alternative function `clue::solve_LSAP` for larger instance.

```
set.seed(42)
n = 3000
cost_mx = matrix(sample(1:1000, n*n, replace = TRUE), nrow = n)

system.time({
  sol_rcpp = RcppHungarian::HungarianSolver(costMatrix = cost_mx)
})

system.time({
  sol_clue = clue::solve_LSAP(x = cost_mx, maximum = FALSE)
})
```