Fix a performance problem with the email address matching filter. The email address regular expression was very general and could match very long pieces of text which were not actually email addresses.

Adding word boundary markers improves this significantly.

See the python xample code here for a test case: <a href="http://bugzilla.gnome.org/show_bug.cgi?id=350015">http://bugzilla.gnome.org/show_bug.cgi?id=350015</a>;