Add random seed argument to resampling functions

Instead of having to set the random seed outside of calls to create resample objects, it would be helpful to a allow passing the seed as an argument. This would help with reproducibility as the seed would be part of the function and could even be preserved as an attribute in an rset object.

For example, something like this:

resample_function(<current arguments>, seed = NULL){
  if(!is.null(seed) && is.numeric(seed)) set.seed(seed)
  ...
}

The default would still allow setting the seed outside the function as currently supported. It would act similar to sklearn's random_state argument.

Start by reviewing the resampling functions and their current seed-handling behavior. Determine the scope of functions affected, then verify that an optional seed supports reproducible resampling while preserving existing external seeding and can be retained on an rset object.