Performance improvements

• Use Regex Marks instead of iteratively matching each token independently. Thanks @nikic for the Feedback!
• Use strstr and substr instead of regexps in some cases
• Remove AbstractTokenDefinition->getRegex() as it is no longer used