Improve error message on attribute that violates one_of constraint

[travissimon]

This commit updates the error message when a one_of constraint is violated to include the value that caused the error. This can make debugging easier when resources have a large number of states.

Contributor checklist

Leave anything that you believe does not apply unchecked.

• I accept the AI Policy, or AI was not used in the creation of this PR.
• Bug fixes include regression tests
• Chores
• Documentation changes
• Features include unit/acceptance tests
• Refactoring
• Update dependencies

I did not create a test case because this happens when the code will not compile and I was not sure how to create a test for that situation =/.

[jimsynz]

Thanks for the PR — the motivation is solid. Showing more information in these error messages definitely helps when debugging resources with many states.

I have a few concerns with the current implementation:

1. The diff calculation shows the wrong thing

all_states -- constraints[:one_of] shows states that exist in the state machine config but are missing from the attribute's one_of. It doesn't show extra values in one_of that aren't valid states. The "unexpected value" framing suggests it's showing values that shouldn't be there, but it's actually showing values that are absent.

2. Grammar nit

has a 'one_of' constraints — should be singular (constraint) or drop the article (has 'one_of' constraints).

3. Inconsistency with the enum error just below

Lines 100–107 handle the same kind of mismatch for Ash.Type.Enum and use an "expected / got" pattern showing both lists. I think matching that pattern here would be clearer and more consistent:

message: """
Expected the attribute #{attribute.name} to have the `one_of` constraint with the following values:
#{inspect(Enum.sort(all_states))}

Got

#{inspect(Enum.sort(constraints[:one_of] || []))}
"""

This shows everything needed to spot the mismatch without computing a diff that might be confusing.

4. "Unexpected value" framing is misleading

This error fires at compile time when the DSL config doesn't match the attribute constraints — it's not a runtime value error. The original "Expected the attribute to have..." phrasing was more accurate about the nature of the problem.

[jimsynz]

Thanks for the contribution! 🚀