Compatibility for resample weights with racing methods

[tjburch]

Finetune's racing functions (tune_race_anova, tune_race_win_loss) internally slice and reorder rsets using restore_rset() and randomize_resamples(). These all are incompatible against the new resample weights:

library(rsample)

set.seed(1)
folds <- vfold_cv(mtcars, v = 5)
weights <- c(1.0, 2.0, 3.0, 4.0, 5.0)
attr(folds, ".resample_weights") <- weights

# restore_rset: slices rows but leaves weights at original length
sliced <- finetune:::restore_rset(folds, 1:3)
length(attr(sliced, ".resample_weights"))
#> [1] 5 -- should be 3

# randomize_resamples: shuffles rows but weights stay in original order
set.seed(42)
shuffled <- finetune:::randomize_resamples(folds)
identical(attr(shuffled, ".resample_weights"), weights)
#> [1] TRUE

This PR updates racing_helpers.R to correctly handle the sample weights:
restore_rset() now subsets .resample_weights[index] alongside row.names, and randomize_resamples() tracks original row positions to reorder weights after the shuffling.