Improve pg_search documentation around tokenizer details

[neilyio]

What
Specifically, we should make clear that search operations like regex, fuzzy find, etc all take place on the tokens that are generated by the index tokenizer. Not necessarily the row data itself.

For example, a username of johnlovemarry might be tokenized as "john love marry" or tokenized as "johnlovemarry", and it would affect how a regex of "(love)" would match it... matching john love marry and not johnlovemarry.

Some examples of how the tokenizers work on simple words would go a long way.