Makes some connection specs less vague

the videnc quiddity was exposing its caps as any which made switcher sometimes try to connect video into audio.

It now exposes video/h264 and video/h265 which is inaccurate (we would need to dynamically determine the caps with a meta sfid and set it to the right cap when we allocate it) but good enough for our purposes.

This also makes the decoder quiddity have 3 different meta swid so that it can tell its real swid if they are writing video audio or text instead of the real swid saying it did all these three things at once.