plots with missing and NaN data

Missing and missing are first class types/values in Julia Base. PythonPlot should simply not plot missing data. It errors in PythonPlot.
```julia
julia> plot([1, missing, 2])
ERROR: Python: ValueError: setting an array element with a sequence.
```

Here is a type translation inspired by one that works for PyCall (https://github.com/JuliaPy/PyPlot.jl/issues/593):
```julia
using PythonPlot
using PythonCall

"Recast julia Arrays with missing as python masked arrays."
function PythonCall.Py(a::Array{Union{T,Missing},N}) where {T,N}
np = PythonCall.pyimport("numpy")
np.ma.masked_invalid(np.array(replace(a, missing => NaN)))
end

# test
plot([1,2,missing,4,5], marker="o")
```

Start with the Julia reproducer `plot([1, 2, missing, 4, 5], marker="o")` and inspect the PythonCall conversion path used by PythonPlot. Compare it with the proposed `PythonCall.Py` masked-array translation, then verify that plots containing `missing` no longer raise the shown ValueError and omit those values.

由索引模型根据 Issue 内容生成。