improved deduping across languages

[missinglink]

following on from #1222 this PR changes two behaviours:

• improved matching across languages
• matching on aliases

improved matching across languages

The previous code would cycle through all the names in item1 and first do a key-check on item2, only if the key also existed in the second item, it would then compare values, if they differed then the two items were considered different.

The issue with that approach is something like this, which is considered different:

var item1 = {
  'name': {
    'default': 'Bern',
    'eng': 'Bern'
  }
};
var item2 = {
  'name': {
    'default': 'Bern',
    'eng': 'Berne'
  }
};

The new code does it differently, instead it first cycles through all the values in item2, comparing them to item1.default, then it cycles through all the values in item1 and compares them to item2.default.

So it's essentially saying "If any of the language values matched the default value of the other item, consider them as a dupe"

[edit] I added another commit which now also considers the user agent language for comparison on top of the default.

matching on aliases

We were only matching on the first item in an array of values, this code uses slightly more CPU but it considers all aliases in deduplication.

[orangejulius]

I tested this one on dev, and it doesn't break any acceptance tests, but I also couldn't find any improvements. I'm guessing this requires some of the other deduping PRs as well.

If you can think of a case that should be improved by using just this PR, let me know

[missinglink]

included in #1230