webrtcsink: NVIDIA V4L2 encoders always require NVMM memory

And if the input is not like that then a corresponding converter must be inserted.

Merge request reports