Fix types in inits function

[MartinuzziFrancesco]

Description

With rcjl 0.11 the minimal inits can also use arrays as input for their weight kwargs. However, in some functions the type Number is still forced (is here)

Expected behavior

weight keywords should also accept arrays as inputs

Minimal Reproducible Example

esn = ESN(input_data, input_size, res_size;
    reservoir=cycle_jumps(;cycle_weight=randn(res_size)),
    input_layer=weighted_init,
    nla_type=NLAT2(),
    rng=rng)

ERROR: TypeError: in keyword argument cycle_weight, expected Number, got a value of type Vector{Float64}
Stacktrace:
 [1] (::WeightInitializers.PartialFunction.Partial{…})(::MersenneTwister, ::Type, ::Vararg{…}; kwargs::@Kwargs{})
   @ WeightInitializers.PartialFunction ~/.julia/packages/WeightInitializers/TEYjX/src/partial.jl:40
 [2] (::WeightInitializers.PartialFunction.Partial{…})(::MersenneTwister, ::Type, ::Vararg{…})
   @ WeightInitializers.PartialFunction ~/.julia/packages/WeightInitializers/TEYjX/src/partial.jl:38
 [3] ESN(train_data::Matrix{…}, in_size::Int64, res_size::Int64; input_layer::typeof(weighted_init), reservoir::WeightInitializers.PartialFunction.Partial{…}, bias::typeof(zeros32), reservoir_driver::RNN{…}, nla_type::NLAT2, states_type::StandardStates, washout::Int64, rng::MersenneTwister, matrix_type::Type)
   @ ReservoirComputing ~/.julia/packages/ReservoirComputing/yPGlg/src/esn/esn.jl:71
 [4] top-level scope
   @ ~/Documents/mcesn/mcesn.jl:32
Some type information was truncated. Use `show(err)` to see complete types.

Potential fixes

Just relax kwargs types and this is solved.

[MartinuzziFrancesco]

Related to this is to be more careful in forcing precision, here gives an error if an array is provided