One way to do it would create a stanfit object from the cmdstan csv output files:
stanfit <- rstan::read_stan_csv(linear.fit.3$output_files())
then you could proceed as you would using RStan.
Another way would be to use the posterior package to get a data frame that you can play with:
library(posterior)
# an array of draws from posterior
draws_array <- linear.fit.3$draws()
# convert to data frame
draws_df <- as_draws_df(draws_array) # or as_draws_matrix() for matrix
The posterior package also can be used in conjunction with bayesplot:
mcmc_hist(linear.fit.3$draws("y_rep"))
See the getting started with cmdstanr documentation for more.