Store some additional tracker data in tracking_map

[ghostwords]

This adds tracking_map, a new Badger storage area. It is similar to snitch_map, but instead of tracker base domains pointing to arrays of site base domains, tracker base domains point to objects keyed by site base domains with arrays of detected tracking types for values. For example:

"arkoselabs.com": {
  "cheaptickets.com": [
    "canvas"
  ],
  "expedia.ca": [
    "canvas"
  ],
  "expedia.co.uk": [
    "canvas"
  ]
},

For now, just two tracking types are being tracked: "canvas" (fingerprinting) and "pixelcookieshare".

Here is a recent 20K site Chrome run: results.zip. Let's compare this/bigger scans to Disconnect's FingerprintingInvasive domains and Tracker Radar fingerprinters.

Should help with #1527 by at least helping us understand the extent of the problem.

Could be followed up by calling out domains that engage in particularly invasive tracking (such as canvas fingeprinting) in the UI (related to #963).

Replaces https://github.com/EFForg/privacybadger/tree/explain-blocking.

[ghostwords]

Also, here is a recent 20K site Chrome run with blocking disabled (TRACKING_THRESHOLD set to 20001): results-no-blocking.zip.

Useful console queries:

Prints the ten most prevalent tracker base domains along with their prevalence (how many site base domains they were seen on):

let sm = badger.storage.snitch_map;
Object.keys(sm._store).map(d => [d, sm._store[d].length]).sort((a, b) => b[1] - a[1]).slice(0, 10).reduce((memo, i) => { memo.push(i[0] + ": " + i[1]); return memo; }, [])

Prints the most prevalent canvas fingerprinters:

let tm = badger.storage.tracking_map;
Object.keys(tm._store).filter(d => Object.keys(tm._store[d]).some(s => tm._store[d][s].includes("canvas"))).map(d => [d, Object.keys(tm._store[d]).length]).sort((a, b) => b[1] - a[1]).slice(0, 25).reduce((memo, i) => { memo.push(i[0] + ": " + i[1]); return memo; }, [])