Allow Naive passing of Julia class_weight dictionary

[tylerjthomas9]

Ref: #63

I have also added Aqua.jl to the tests, and fixed some compt errors that it identified.

I introduced a function, _prepare_param, that handles preparing parameters being passed to the python constructor. Right now, it does nothing unless a Dictionary is passed. I tried to make it generic for future extensions to other types as needed.

[codecov]

Codecov Report

All modified and coverable lines are covered by tests ✅

Comparison is base (530ce27) 94.49% compared to head (75f718a) 94.60%.

Additional details and impacted files

@@            Coverage Diff             @@
##              dev      #64      +/-   ##
==========================================
+ Coverage   94.49%   94.60%   +0.10%     
==========================================
  Files          14       14              
  Lines         309      315       +6     
==========================================
+ Hits          292      298       +6     
  Misses         17       17              

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

[ablaom]

Looks good to me. Thanks for this valuable contribution, @tylerjthomas9 🙏🏾

Could you have a look at the 1.6 fails, please?

[tylerjthomas9]

The 1.6 failures were due to Aqua.jl saying there is a method ambiguity in MLJModelInterface.jl. I have disabled checking them, for now. Something funky is going on with the nightly build. Do you think its worth investigating before merging?

Here is the 1.6 ambiguity from Aqua.jl:

1 ambiguities found
Ambiguity #1
in(x::MLJModelInterface.MLJType, itr::AbstractVector{T} where T) in MLJModelInterface at /Users/runner/.julia/packages/MLJModelInterface/ieFuy/src/equality.jl:158
in(x::T, r::AbstractRange{T}) where T in Base at range.jl:1088

Possible fix, define
  in(::T, ::AbstractRange{T}) where T<:MLJModelInterface.MLJType

[ablaom]

I've flagged MLJModelInterface.jl issue found by Aqua. I'm happy not to dig into the nightly issue at this time. So go ahead and merge/tag when ready.

[tylerjthomas9]

I am going to hold off on tagging because I want to look into the issues with fitting classifiers with a single class here and with CatBoost.jl, then just push a release with both of them.