fix($location): ensure $locationChangeStart broadcast during $digest

[caitp]

I'm not totally convinced this is a necessary change (see comments on #5118).

However, feedback valuable. I'll amend this message if it becomes clear that
it's a worthwhile change.

Closes #5118
Closes #4989

[IgorMinar]

this is also related to #4989

[caitp]

The issue is that the $browser calls fireUrlChange iterating over each of the listeners without wrapping in a $digest --- the effect is that the event is emitted outside of a digest.

The reason I say this change is sort of questionable is because it's not clear that it really matters in any significant fashion, but I suppose it's better to not force people to check $$phase

So we do fire a digest later, but this is after the event is already handled, and it's all unrelated to the if statement

[caitp]

Hmm, I guess maybe I favour #5089 over this, after re-thinking it

[IgorMinar]

I'm thinking that we should do both this and #5089.

[caitp]

I see ... well in any case, both PRs are missing tests, I'll write some up tomorrow I guess

[IgorMinar]

I'll add tests and merge them both today. I want to be done with this issue :)

[caitp]

fair enough, PS happy newyear \o/

[IgorMinar]

I just realized that by moving the broadcast into evalAsync, we will always execute it within a $digest phase, so this PR is not necessary.

I think we'll get rid of the $phase check and double $digest error in 1.3 (we talked about in the office today and I think I know how to do it), so with that change this $phase check will disappear and we'll always wrap the whole thing into $apply without causing double digest error.

[IgorMinar]

oh, yeah. happy new year! :)