audioencoder: Baseclass should call set_format() before propose_allocation()

Submitted by Nicolas Dufresne @ndufresne

Link to original bug (#732874)

Description

This was introduce during porting https://bugzilla.gnome.org/show_bug.cgi?id=680614.

The downside is that it makes it counter intuitive, as propose_allocation() should happen after set_format(). To implement propose_allocation(), encoder have to get the caps from the pad, and then most likely parse these caps into GstVideoInfo structure. Without this patch, it would simply have cached the input_state, and read the information from there.

I do think this is in the ABI now, but would at least be nice to revisit for 2.0.