Document pattern of tidying draws before spread/gather_draws

[mjskay]

With new optimizations (spurred on by #144) spread_draws and gather_draws should be fast in most cases even with large numbers of parameters. However tidy_draws can still be somewhat slow when there are lots of parameters.

One way to deal with this is call tidy_draws() once ahead of time and only use pre-tidied draws with spread_draws() and gather_draws(). Could be useful to document this pattern.

[mjskay]

can also mention shredder as an option