Improve support (or documentation) for weighted draws

?weight_draws says Add weights to draws objects, with one weight per draw, for use in subsequent weighting operations. Currently the only supported subsequent operation is resample_draws(), but the documentation doesn't make it clear. Ideally rvars support would be extended to support weighted draws, but if that is infeasible, the documentation should be clarified.

Example:

> x = rvar_rng(rnorm, 4, mean = 1:4, sd = 2)
> w=runif(4000)

weight_draws() is not defined for draws_rvars:

> xw = weight_draws(x, w, log=FALSE)
Error in UseMethod("weight_draws") : 
  no applicable method for 'weight_draws' applied to an object of class "c('rvar', 'vctrs_vctr', 'list')"

There is no error if weighted draws_df is converted to draws_rvars:

> xw = as_draws_rvars(weight_draws(as_draws_df(x), w, log=FALSE))

The default print method gives the unweighted result

> xw
# A draws_rvars: 4000 iterations, 1 chains, and 1 variables
$x: rvar<4000>[4] mean ± sd:
[1] 0.96 ± 2  1.95 ± 2  3.02 ± 2  4.05 ± 2 

and mean fails

> mean(xw)
[1] NA
Warning message:
In mean.default(xw) : argument is not numeric or logical: returning NA

Start with the weight_draws(), resample_draws(), and as_draws_rvars() entry points, then inspect how draws_rvars objects handle weighted draws and summary operations. Done means either weighted draws_rvars support works, including mean(), or the documentation clearly states the supported limitation and conversion path.