Optimisation: faster extraction of META tags

[jnioche]

The utility classes RefreshTag and RobotsTags both use XPATH to retrieve META tags. They currently do so by looking for //META which is inefficient as it searches everywhere in the document. These 2 methods can take up to 18% of the processing time for JSoupParserBolt and 16% of the overall CPU.

Instead, we can use a more constraining XPATH which will look only into /HTML/HEAD or /HTML/BODY, the latter is not the recommended variant but can be found in the wild.

[jnioche]

Profiling after the change doesn't show a significant impact on RobotsTags.extractMetaTags but RefreshTag takes only 1/2 the time it used to. This represents 14% of the processing time for JSoupParserBolt and 11% of the overall CPU.