ScientificTypes.TRAIT_GIVEN_FUNCTION not updating; info(rms) broken

[ablaom]

julia> info(rms)
ERROR: MethodError: no method matching info(::MLJBase.RMS, ::Val{:other})
Closest candidates are:
  info(::Any) at /Users/anthony/Dropbox/Julia7/MLJ/MLJBase/src/MLJBase.jl:188
  info(::Any, ::Val{:measure}) at /Users/anthony/Dropbox/Julia7/MLJ/MLJBase/src/measures.jl:91
Stacktrace:
 [1] info(::MLJBase.RMS) at /Users/anthony/Dropbox/Julia7/MLJ/MLJBase/src/MLJBase.jl:188
 [2] top-level scope at none:0

Remedy: put ScientificTypes.TRAIT_GIVEN_FUNCTION updates back into init

(baby went out with the bathwater in last patch 😢)

[tlienart]

if we can we should avoid the use of init (hacky), can this not be achieved otherwise?

(also I apologise given I think it's me who modified this; was this not tested then?)

[ablaom]

Sorry, I did not see this comment before proceeding.

There's no simple way to avoid the init. One package is updating a global variable defined in another package.

I don't think init is necessarily hacky. The issues we have had in the past are to do with eval in __init__. That is not the case here.

If it is really an issue, I will look into an alternative approach. Can you confirm this really is an issue?

I've included a test that catches the trip-up.