BUG: fallback to default on best_match

[wasade]

...avoid a possible None for language

[dhakim87]

Definitely approve of setting default on best_match and tests for invalid tags, but

I don't understand why normalize is necessary. All calls are coming in through microsetta-interface, which sets language tag to whatever we specify. Where the heck did a tag with the wrong casing come from???

[wasade]

Worth having the normalize to be defensive or should it be dropped?

[dhakim87]

I'd rather remove the normalize if we can track down who is sending up "en_us", "en_us" is used as a key sent to babel so being passed up may mean we show the wrong language consent form or something because we forgot to do the lookup.