Add git lfs tracked models and change default model paths of filter set #1 to the new models. Add training scripts of JVET-AC0089.

Renjie Chang authored 2 years ago and Liqiang Wang committed 2 years ago

Integrate JVET-AC0196 (EE1-2.2: GOP Level Adaptive Resampling with CNN-based Super Resolution), only inference code part

Both the floating point and integer SADL models for luma intra are updated.
Filtering changed so that partitioning input is no longer input to the network.
Full JVET-AB0053 title: EE1-1.2: NN intra model without attention and partitioning strength