Fastmri

This update includes a suite of bug fixes.

- Fix for `apply_mask` function with single coil data (PR 255)
- New filter function feature for filtering contrasts with a `SliceDataset` (PR 257).
- Fix for `AnnotatedSliceDataset` annotation versions (PR 276).
- Fix typecast in varnet (PR 303).
- Fix annotation metadata extraction for `AnnotatedSliceDataset` (PR 310).
- Fix calculation of y location in `AnnotatedSliceDataset` (PR 311).

Contributors: zaccharieramzi NikolasMorshuis ad12

This release of fastMRI includes PyTorch Lightning modules for the paper [On learning adaptive acquisition policies for undersampled multi-coil MRI reconstruction](https://arxiv.org/pdf/2203.16392.pdf) by Bakker et al., to be presented at the upcoming MIDL conference. A list of changes is below:
[](url)
- Add Adaptive VarNet implementation (PR 205)
- Compatibility for fastMRI+ in SliceDataset (PR 192) - currently in beta.
- Masking API updates (PR 167).
- Various smaller bug fixes and improvements.

Contributors:
Timsey luisenp michaldrozdzal

- Includes a fix for packaging that removes `fastmri_examples` from distributed sources. (145).
- Updates the FFT API to allow the user to adjust normalizations (146).

This release does not introduce any major changes - it merely includes metadata for publishing the project on PyPI. The creation of this release also tests the GitHub release action.