Fix Resampler update bug

[ablaom]

This PR fixes a problem when the model wrapped by a Resampler instance is mutated to one of different prediction type. I ran into this while working on JuliaAI/MLJTuning.jl#201.

using MLJBase, MLJModels
KNNClassifier = @load KNNClassifier

knn = KNNClassifier()
cnst =DeterministicConstantClassifier()

X, y = make_blobs()

resampler = MLJBase.Resampler(model=knn, measure=accuracy, operation=nothing)
mach = machine(resampler, X, y) |> fit!

resampler.model = cnst
fit!(mach)
# [ Info: Updating machine(Resampler(model = DeterministicConstantClassifier(), …), …).
# ERROR: MethodError: no method matching iterate(::CategoricalArrays.CategoricalValue{Int64, UInt32})

# Closest candidates are:
#   iterate(::LibGit2.GitBranchIter)
#    @ LibGit2 /Applications/Julia-1.10.app/Contents/Resources/julia/share/julia/stdlib/v1.10/LibGit2/src/reference.jl:343
#   iterate(::LibGit2.GitBranchIter, ::Any)
#    @ LibGit2 /Applications/Julia-1.10.app/Contents/Resources/julia/share/julia/stdlib/v1.10/LibGit2/src/reference.jl:343
#   iterate(::Parameters.Lines)
#    @ Parameters ~/.julia/packages/Parameters/MK0O4/src/Parameters.jl:78
#   ...

# Stacktrace:
#   [1] isempty(itr::CategoricalArrays.CategoricalValue{Int64, UInt32})
#     @ Base ./essentials.jl:952
#   [2] mode(a::CategoricalArrays.CategoricalValue{Int64, UInt32})
#     @ StatsBase ~/.julia/packages/StatsBase/WLz8A/src/scalarstats.jl:112
#   [3] _broadcast_getindex_evalf
#     @ ./broadcast.jl:709 [inlined]
#   [4] _broadcast_getindex
#     @ ./broadcast.jl:682 [inlined]
#   [5] getindex
#     @ ./broadcast.jl:636 [inlined]
#   [6] copy
#     @ ./broadcast.jl:942 [inlined]
#   [7] materialize
#     @ ./broadcast.jl:903 [inlined]
#   [8] predict_mode(m::DeterministicConstantClassifier, fitresult::CategoricalArrays.CategoricalValue{…}, Xnew::@NamedTuple{…})                                
#     @ MLJBase ~/MLJ/MLJBase/src/interface/model_api.jl:11
#   [9] predict_mode(mach::Machine{DeterministicConstantClassifier, true}; rows::UnitRange{Int64})                                                                                    
#     @ MLJBase ~/MLJ/MLJBase/src/operations.jl:86
#  [10] predict_mode
#     @ ~/MLJ/MLJBase/src/operations.jl:82 [inlined]

[codecov]

Codecov Report

All modified and coverable lines are covered by tests ✅

Comparison is base (193ad67) 87.95% compared to head (9742d44) 87.99%.

Additional details and impacted files

@@            Coverage Diff             @@
##              dev     #954      +/-   ##
==========================================
+ Coverage   87.95%   87.99%   +0.03%     
==========================================
  Files          28       28              
  Lines        2540     2540              
==========================================
+ Hits         2234     2235       +1     
+ Misses        306      305       -1     

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