Expectations about types in `ratios` seem too strict

[ablaom]

julia> oversampler = SMOTE(rng=stable_rng(), ratios=Dict("negative"=>1, "positive"=>0.5))
ERROR: MethodError: no method matching var"#SMOTE#3"(::Int64, ::Dict{String, Real}, ::StableRNGs.LehmerRNG, ::Bool, ::Type{Imbalance.MLJ.SMOTE})
The function `#SMOTE#3` exists, but no method is defined for this combination of argument types.

Closest candidates are:
  var"#SMOTE#3"(::Integer, ::Union{Nothing, AbstractFloat, Dict{T, <:AbstractFloat}}, ::Union{Integer, Random.AbstractRNG}, ::Bool, ::Type{Imbalance.MLJ.SMOTE}) where T        
   @ Imbalance ~/.julia/packages/Imbalance/LFVkH/src/oversampling_methods/smote/interface_mlj.jl:30                                                                         

Stacktrace:
 [1] top-level scope
   @ REPL[343]:1

This error message is not too helpful for beginners. The problem is that "negative"=>1 needs to be "negative"=>1.0. We ought to just allow mixed data here, no?