Proposal: use of Singleton pattern for Plottertype

[jpthiele]

I think we can make better use of dispatch if we employ the Singleton pattern as described in
https://www.packtpub.com/en-us/product/hands-on-design-patterns-and-best-practices-with-julia-9781838646615
This is practically identical to Val{T} but can be used with a custom more obviously named type as well.

This would greatly simplify some of the functions in dispatch.jl, e.g. plottertype
This is an example of how it could look like

abstract type AbstractPlotterType end
struct PlotterType{T} <: AbstractPlotterType end #The singleton type

plottertype(m::Union{Module,Nothing}) = PlotterType{nameof(m)}() #Return singleton 
plottername(::Plottertype{T}) where {T} = String(T)

const PythonPlotType = PlotterType{:PythonPlot} #Shorthand for compatibility
const PyPlotType = PlotterType{:PyPlot} 
const AbstractPythonPlotterType = Union{PyPlotType,PythonPlotType} #makes PyPlotType <: AbstractPythonPlotterType true

We can verify things work as expected (in the small test)

PyPlotType === PlotterType{:PyPlot} # returns true
PyPlotType() === PlotterType{:PyPlot}() # returns true

p1 = PyPlotType()
p2 = PyPlotType()
p1 === p2 # returns true since PyPlotType is a singleton


function foo(a::T) where {T<:AbstractPythonPlotterType} 
       @info "Yay"
end
foo(PyPlotType()) # Outputs [ Info: Yay

So I think this would pretty much be a drop-in replacement but before I go through all of dispatch.jl and prepare a PR I wanted to suggest it first to see what you think.

[jpthiele]

If I don't overlook something then the functions like ismakie are only used in the custom plot example and internally, correct?
The example could be solved using dispatch again with private functions within plotting_custom

(not tested would need the full PR!)

function customscatter!(ax,Plotter) = scatterpoints!(ax,plottertype(Plotter))
function customscatter!(ax,::MakiePlotType) = Makie.scatter!(ax,rand(10),rand(10),fill(0.1,10); color=:blue,markersize=20)
function customscatter!(ax,::PyPlotType) = ax.scatter(rand(10),rand(10), s= 500)
function customscatter!(ax,::PlotsType) = Plots.scatter!(ax,rand(10),rand(10); color= :blue, markersize=10, label=nothing)

function plotting_custom(; Plotter= default_plotter(), kwargs...) 
   vis = GridVisualizer(;Plotter=Plotter)
   grid = grid2d()
   gridplot!(vis,grid)
   
   customplot!(vis) do ax
     customscatter!(ax,Plotter)
   end

It looks like this would make the whole thing more extendable as well.