[jl-wynen]

Fixes #120

(Draft for now because I am using CI to run the tests)

I considered a couple of alternatives:

1. Override NeXusClass[snx.NXsource] as we used to do for BIFROST. But as noted in Consider loading source from NXmoderator #120 (comment), this breaks code that expects Component[snx.NXsource] to be an actual NXsource.
2. Override compute_position(transform: NeXusTransformation[snx.NXmoderator) -> Position[snx.NXsource] This would avoid the problem in point 1, but would still be strange because it hides the actual NXsource and could lead to confusion.
3. Explicitly request moderator_position: Position[snx.NXmoderator] in all providers. This would be clean but would lock us in to using the moderator everywhere.
4. (chosen) Make a distinct NeutronSourcePosition type. This allows us to provide positions from either the moderator or source, depending on what file we read (e.g., ESS vs McStas)

• I opted to raise an exception whenever someone requests Position[snx.NXsource] because that is an easy mistake to make but hard to spot.

@YooSunYoung The NMX workflow tests fail because the test file does not contain a moderator. Are the tests designed for actual ESS files or for simulation data? In the former case, we should update the file. In the latter, we should update the workflow to load NXsource.

[nvaytet]

Suggested change

	"insert your providder to overridethe one that produced this error."
	"insert your provider to override the one that produced this error."