Reports view: split firefox/thunderbird strings

[pascalchevrel]

The Reports view gives for desktop an inaccurate count of missing/translated strings because it compares it to the number of strings in en-US which includes firefox, thunderbird ans seamonkey. A locale like Afrikaans which only does Firefox will get a 20% completion while it would be a much higher number if we were counting only Firefox strings.

I am not sure yet what approach to take as there are many possible (extract strings in separate files per app in python, analyse strings at run time with the entity names, creacte cache files splitting per app post extraction...)