Make phutil_tag() / phutil_escape_html() 10-15% faster

Summary:
These numbers are a bit rough, but they seem to make a difference. Total weight of my test page (Maniphest task list) dropped from ~194ms to ~170ms. Time spent in phutil_tag() dropped from 105-110ms to 90ms.

This also cleans up the call graph for xhprof a bit (e.g., removes 1,300 calls to array_map()). Generally, foreach(...) is faster than array_map(...) (at least for small inputs, and at least on Zend PHP) as we avoid paying for function call overhead.

Test Plan: Loaded various pages. There should be no functional changes.

Reviewers: btrahan, chad, vrana

Reviewed By: chad

CC: aran

Differential Revision: https://secure.phabricator.com/D6183