CMRxRecon challenge 2023 participation scripts

[georgeyiasemis]

This PR incrorporates code used for DIRECT team participation at the CMRxRecon challenge 2023:

• CMRxReconDataset which allows for 2D or 3D data loading.
• Cartesian subsampling functions (similar to :class:FastMRI, but instead of center_fractions representing the fraction of center lines to the original size, here, it represents the actual number of center lines).
• Training and inference scripts (yaml, lst files) for reproducing our experiments.

Also it introduces improvements to documentation and typing for edited files.

[codecov]

Codecov Report

Attention: Patch coverage is 88.01498% with 32 lines in your changes are missing coverage. Please review.

Project coverage is 84.80%. Comparing base (79543f0) to head (99d4b7a).

❗ Current head 99d4b7a differs from pull request most recent head 4caa1fb. Consider uploading reports for the commit 4caa1fb to get more accurate results

Files	Patch %	Lines
direct/data/datasets.py	82.66%	26 Missing ⚠️
direct/common/subsample.py	94.17%	6 Missing ⚠️

Additional details and impacted files

@@            Coverage Diff             @@
##             main     #274      +/-   ##
==========================================
+ Coverage   84.76%   84.80%   +0.03%     
==========================================
  Files          96       96              
  Lines        6963     7185     +222     
==========================================
+ Hits         5902     6093     +191     
- Misses       1061     1092      +31     

☔ View full report in Codecov by Sentry.
📢 Have feedback on the report? Share it here.

[jonasteuwen]

Some comments, in general:

• Might it be possible to deduplicate or define more abstract base classes?
• Use pathlib
• Avoid randomness in the dataset classes

[jonasteuwen]

Looks good to me