Remove special handling for the $sitekey option

[trev]

Background

Currently the $sitekey option is handled separately from any other filter options. The background here is that it was originally meant to whitelist everything on particular websites - so I implemented this special solution in order to avoid impacting the performance negatively. In addition to requiring more code, distinguishing between individual requests isn't possible and only one filter is respected for each key.

What to change

Remove the special handling for sitekeys in lib/matcher.js - they should be handled in exactly the same way as the domain option. That means that the filters are matched by the pattern part first, other options are checked later. This means that the sitekey needs to be added as a parameter for Mather.matchesAny() and RegExpFilter.matches(). The latter can check sitekey matches trivially.

The complication here is: a valid sitekey should apply to the current page and any pages below it in the frame hierarchy. E.g. consider the following structure:

top
|- frame1 (sitekey)
   |- frame2

We should consider the sitekey both when checking whether frame1 and frame2 are whitelisted. Also, any request triggered by frame1 or frame2 should be checked with the sitekey considered. top on the other hand is above the sitekey, so there it shouldn't be considered.

There is also the unlikely scenario that there are multiple valid sitekeys in the hierarchy. IMHO we don't need to overcomplicate things here - whichever sitekey is closer should be used:

top (sitekey1)
|- frame1 (sitekey2)
   |- frame2

Here sitekey1 should be checked when determining whether top is whitelisted, also for any requests made by top. For frame1 and frame2 it is being overwritten with sitekey2 however, so sitekey2 should be used for these frames instead.

[greiner]

Fixed: ​https://hg.adblockplus.org/adblockplus/rev/026649336216