Add entropy and collision probability estimation.

[lcrh]

This provides functions to estimate entropy and collision probability of IDs generated via the generate call. These estimates will be accurate if no func_words (such as 'uuid/') are used. In the presence of functions, sampling is used to estimate entropy of the function. As the number of samples increases this will approach an upper bound on the entropy (which will underestimate collision probability). If the function is not a typical ID function, then the upper bound may be a lot higher than the true entropy, and hence collision probability estimates will be too low.

[lcrh]

Quick ping on this pull request.