Add other physics-informed deep learning methods with pretrained models

[Andrewwango]

Some models to add:

• PnP/RED with pretrained denoisers (e.g. see this demo) - we would need to find a pretrained denoiser from the internet somewhere
• Diffusion sampling with pretrained denoisers, e.g. DiffPIR/DDRM (e.g. see this demo)
• Pretrained unrolled models such as the unrolling of HQS (i.e. MoDL)
• Other pretrained models such as VarNet. In this specific case we could get one from the FastMRI examples (although note they don't seem to have one pretrained for singlecoil data (see here))