`mean`'s dry run

[thevolatilebit]

I used to assume that mean(f, itr) calls f once for each of the elements of itr.

However, there is an (extra) dry run which

• may lead to unwanted behaviour for fs that mutate the state,
• may negatively influence the overall performance (my primary motivation).

In Slack's #helpdesk channel, it was also noted that there could be some issues with weirdly indexed arrays.

For example,

g3 = 0; mean(i -> (global g3 += 1; 1), 1:1)

results in g3 == 2.

Moreover, this behaviour is not documented.

[fredrikekre]

See #49