Clearer distinguish length and max

[sbfnk]

Fixes #410

Also updates examples and tests to have their max value reduced by 1 (as the previous interpretation of max parameters was to be the length).

This required a slight change to what dist_skel to return NA if asking for the probability mass of a value that is greater than the max.

[github-actions]

This is how benchmark results would change (along with a 95% confidence interval in relative change) if b9de896 is merged into main:

•   :ballot_box_with_check:default: 44.4s -> 45.9s [-9.42%, +16.03%]
•   :ballot_box_with_check:no_delays: 45.1s -> 47.6s [-5.9%, +16.64%]
•   :ballot_box_with_check:random_walk: 14.1s -> 15s [-1.93%, +15.11%]
•   :ballot_box_with_check:stationary: 27.8s -> 27.9s [-6.59%, +7.83%]
•   :ballot_box_with_check:uncertain: 1.11m -> 1.16m [-7.13%, +17.66%]
  Further explanation regarding interpretation and methodology can be found in the documentation.

[sbfnk]

LGTM (self-review).

@seabbs would you like to review or happy for it to be merged as is?

[seabbs]

All looks good to me.