PET / MRI entanglement broken, spooky action at a distance not observed

[effigies]

BIDS states that NonlinearGradientCorrection is "RECOMMENDED, but REQUIRED if PET data are present".

This is not implemented in the legacy validator:

https://github.com/search?q=repo%3Abids-standard%2Fbids-validator+NonlinearGradientCorrection&type=code

Nor, in fact, is it actually followed in examples:

https://github.com/search?q=repo%3Abids-standard%2Fbids-examples+NonlinearGradientCorrection&type=code

There is a mis-capitalized version of it in one example.

xref #1229

cc @bendhouseart