DWI derivatives: Make tensor a data representation

[Lestropie]

Related to #73.

In looking over the contents of the ISMRM DSGW abstract producing a poster, this came across as erroneous. While having an arbitrary "OrientationRepresentation" = "param" potentially gives some capacity for supporting exotic models not yet integrated in the specification, it seems to me now that it's not appropriate to use such a mechanism for the tensor. While we think of the diffusion tensor as a "model", it's actually a data representation that is theoretically extensible to other models (eg. multi-tensor models where you might have multiple such images; discrete mixture models where you may have one or more components that are stored as tensors; non-DWI modalities that also have anisotropy information per voxel that may in the future make use of this orientation representation construct).

[codecov]

Codecov Report

All modified and coverable lines are covered by tests ✅

Project coverage is 88.53%. Comparing base (dd8fe32) to head (eddad22).
Report is 248 commits behind head on bep-016.

❗ Current head eddad22 differs from pull request most recent head c6e50fc. Consider uploading reports for the commit c6e50fc to get more accurate results

Additional details and impacted files

@@             Coverage Diff             @@
##           bep-016      #74      +/-   ##
===========================================
+ Coverage    87.88%   88.53%   +0.64%     
===========================================
  Files           14        6       -8     
  Lines         1279     1055     -224     
===========================================
- Hits          1124      934     -190     
+ Misses         155      121      -34     

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