Use angular's parser to analyze expression instead of Regex about angular-gettext-tools

[kvantetore]

I find this approach very interesting. The current regex approach is very limited, and most of the advantage of angular-gettext is the ability to automatically extract translation strings, so improving that aspect seems very important to me.

It's a pretty funky hack to get access to the angular parser. Glancing on parse.js, it looks like it's only dependent on $filter, and even that looks like its using $filter mostly to determine if an expression is constant or not (which we don't really need). So I guess it should be readily mockable.

[kvantetore]

In kvantetore/angular-gettext-tools@7e5e520, I've changed the parser from regex based to the angular expression parser based on the work by @yankee42. I can format a proper pull request if there is interest in folding into the mainline.

Instead of looking for interpolation expressions globally in the html template, the new approach is to parse the attributes and text-tags of each node separately. This has caused a minor regression in the test suite, namely in "extract/Extracts strings from non-delimited attribute", the line number of the second string is reported to be on line 3 instead of 5