Improve linkcheck performance (2x+) getting rid of jtidy dependency via regexps

At the moment, the linkcheck plugin uses jtidy and xpath for retreiving all links. IMHO regexps would work much faster/better than jtidy-xpath combination.

The following regexp would be a replacement for the xpath expressions:

<(?>link|a|img|script)[^>]?(?>href|src)\s?=\s*?[\"'](.?)[\"'][^>]?

All tests pass with this regexp and in project ws-jaxme I am getting these results for maven-linkcheck-plugin:clearcache maven-linkcheck-plugin:report-real:

with jtidy/xpath: Total time: 2 minutes 43 seconds
with regexps: Total time: 1 minutes 10 seconds

I am sure some regexp guru can improve the performance of this.

I have a question, though. Are mailto links supposed to count as checkable? IMO no.

PD: Also, IMO the createDocument method from LinkCheck should be on a try finally block.