Error when fitting with CUDA not functional (adding conditional to fix)

[pat-alt]

Finally having a first go at adding an extension to CounterfactualExplanations.jl and noticed that trying to fit a model when CUDA is not functional throws an error:

julia> using CounterfactualExplanations
       using MLJBase
       using NeuroTreeModels
       using TaijaData
       
       m = NeuroTreeRegressor(; depth=5, nrounds=10)
       X, y = @load_boston
       mach = machine(m, X, y) 
       MLJBase.fit!(mach)
┌ Warning: Package cuDNN not found in current path.
│ - Run `import Pkg; Pkg.add("cuDNN")` to install the cuDNN package, then restart julia.
│ - If cuDNN is not installed, some Flux functionalities will not be available when running on the GPU.
└ @ FluxCUDAExt ~/.julia/packages/Flux/Wz6D4/ext/FluxCUDAExt/FluxCUDAExt.jl:57
[ Info: Precompiling NeuroTreeModels [1db4e0a5-a364-4b0c-897c-2bd5a4a3a1f2]
┌ Warning: Package cuDNN not found in current path.
│ - Run `import Pkg; Pkg.add("cuDNN")` to install the cuDNN package, then restart julia.
│ - If cuDNN is not installed, some Flux functionalities will not be available when running on the GPU.
└ @ FluxCUDAExt ~/.julia/packages/Flux/Wz6D4/ext/FluxCUDAExt/FluxCUDAExt.jl:57
WARNING: Method definition supports_weights(Type{var"#s27"} where var"#s27"<:NeuroTreeModels.NeuroTreeRegressor) in module NeuroTreeModels at /Users/paltmeyer/.julia/packages/NeuroTreeModels/DXITM/src/MLJ.jl:71 overwritten at /Users/paltmeyer/.julia/packages/MLJModelInterface/ihpHk/src/metadata_utils.jl:62.
ERROR: Method overwriting is not permitted during Module precompilation. Use `__precompile__(false)` to opt-out of precompilation.
[ Info: Skipping precompilation since __precompile__(false). Importing NeuroTreeModels [1db4e0a5-a364-4b0c-897c-2bd5a4a3a1f2].
[ Info: Precompiling NeuroTreeExt [3b9b29e0-2f41-503c-9947-dc8035483bb2]
┌ Warning: Package cuDNN not found in current path.
│ - Run `import Pkg; Pkg.add("cuDNN")` to install the cuDNN package, then restart julia.
│ - If cuDNN is not installed, some Flux functionalities will not be available when running on the GPU.
└ @ FluxCUDAExt ~/.julia/packages/Flux/Wz6D4/ext/FluxCUDAExt/FluxCUDAExt.jl:57
┌ Warning: Module NeuroTreeModels with build ID ffffffff-ffff-ffff-0001-d43218202e04 is missing from the cache.
│ This may mean NeuroTreeModels [1db4e0a5-a364-4b0c-897c-2bd5a4a3a1f2] does not support precompilation but is imported by a module that does.
└ @ Base loading.jl:1948
[ Info: Skipping precompilation since __precompile__(false). Importing NeuroTreeExt [3b9b29e0-2f41-503c-9947-dc8035483bb2].
[ Info: Following 14 arguments were not provided and will be set to default: batchsize, MLE_tree_split, gpuID, ntrees, hidden_size, wd, actA, loss, rng, init_scale, device, outsize, lr, stack_size.
[ Info: Training machine(NeuroTreeRegressor(loss = mse, …), …).
[ Info: ["Crim", "Zn", "Indus", "NOx", "Rm", "Age", "Dis", "Rad", "Tax", "PTRatio", "Black", "LStat"]
┌ Error: Problem fitting the machine machine(NeuroTreeRegressor(loss = mse, …), …). 
└ @ MLJBase ~/.julia/packages/MLJBase/iIhiI/src/machines.jl:683
[ Info: Running type checks... 
[ Info: Type checks okay. 
ERROR: CUDA driver not found
Stacktrace:
  [1] error(s::String)
    @ Base ./error.jl:35
  [2] functional
    @ ~/.julia/packages/CUDA/htRwP/src/initialization.jl:24 [inlined]
  [3] task_local_state!()
    @ CUDA ~/.julia/packages/CUDA/htRwP/lib/cudadrv/state.jl:77
  [4] device
    @ ~/.julia/packages/CUDA/htRwP/lib/cudadrv/state.jl:191 [inlined]
  [5] reclaim(sz::Int64)
    @ CUDA ~/.julia/packages/CUDA/htRwP/src/pool.jl:532
  [6] reclaim
    @ ~/.julia/packages/CUDA/htRwP/src/pool.jl:532 [inlined]
  [7] fit_iter!(m::NeuroTreeModels.NeuroTreeModel{…}, cache::@NamedTuple{…})
    @ NeuroTreeModels ~/.julia/packages/NeuroTreeModels/DXITM/src/fit.jl:149
  [8] fit(model::NeuroTreeRegressor, verbosity::Int64, A::@NamedTuple{…}, y::SubArray{…}, w::Nothing)
    @ NeuroTreeModels ~/.julia/packages/NeuroTreeModels/DXITM/src/MLJ.jl:28
  [9] fit(model::NeuroTreeRegressor, verbosity::Int64, A::@NamedTuple{…}, y::SubArray{…})
    @ NeuroTreeModels ~/.julia/packages/NeuroTreeModels/DXITM/src/MLJ.jl:8
 [10] fit_only!(mach::Machine{NeuroTreeRegressor, true}; rows::Nothing, verbosity::Int64, force::Bool, composite::Nothing)
    @ MLJBase ~/.julia/packages/MLJBase/iIhiI/src/machines.jl:681
 [11] fit_only!
    @ ~/.julia/packages/MLJBase/iIhiI/src/machines.jl:607 [inlined]
 [12] #fit!#63
    @ ~/.julia/packages/MLJBase/iIhiI/src/machines.jl:778 [inlined]
 [13] fit!(mach::Machine{NeuroTreeRegressor, true})
    @ MLJBase ~/.julia/packages/MLJBase/iIhiI/src/machines.jl:775
 [14] top-level scope
    @ ~/code/CounterfactualExplanations.jl/test/models/neurotree.jl:9
Some type information was truncated. Use `show(err)` to see complete types.

This PR just adds a conditional statement to avoid this error.