Edge weights not properly documented for `GNNHeteroGraph`s (and implement new function to add new edge weights?)

I was experimenting with edge weights for GNNHeteroGraphs, but I had to dig deep in the documentation (and the docstrings from the source code) to see how edge weights work for heterographs. It'll be nice to add some examples to the documentation page for heterographs covering all operations that one can perform.

Also, it seems like there's no nice function to add new edges with weights to existing graphs. For instance, suppose I create a heterograph with some edge weights:

julia> using GraphNeuralNetworks
julia> g_het = GNNHeteroGraph((:user, :rate, :movie) => ([1,1,2,3], [7,13,5,7], [0.1, 0.2, 0.3, 0.4]))
GNNHeteroGraph:
  num_nodes: Dict(:movie => 13, :user => 3)
  num_edges: Dict((:user, :rate, :movie) => 4)

julia> get_edge_weight(g_het, (:user, :rate, :movie))          # gives us the edge weights
4-element Vector{Float64}:
 0.1
 0.2
 0.3
 0.4

But I can't add new edges with edge weights. Ideally, something like the following will be great:

julia> g_het = add_edges(g_het, (:user, :like, :actor) => ([1, 2], [3, 4], [0.5, 0.6]));          # no method supporting this currently

Also, this leads to some weird behavior; I can't even add edges without weights to g_het:

julia> g_het = add_edges(g_het, (:user, :like, :actor) => ([1, 2], [3, 4]))
ERROR: MethodError: Cannot `convert` an object of type Nothing to an object of type Vector{Float64}

Closest candidates are:
  convert(::Type{T}, ::LinearAlgebra.Factorization) where T<:AbstractArray
   @ LinearAlgebra ~/julia-1.9.0/share/julia/stdlib/v1.9/LinearAlgebra/src/factorization.jl:59
  convert(::Type{Array{T, N}}, ::StaticArraysCore.SizedArray{S, T, N, N, Array{T, N}}) where {S, T, N}
   @ StaticArrays ~/.julia/packages/StaticArrays/9KYrc/src/SizedArray.jl:88
  convert(::Type{Array{T, N}}, ::StaticArraysCore.SizedArray{S, T, N, M, TData} where {M, TData<:AbstractArray{T, M}}) where {T, S, N}
   @ StaticArrays ~/.julia/packages/StaticArrays/9KYrc/src/SizedArray.jl:82
  ...

Stacktrace:
 [1] cvt1
   @ ./essentials.jl:418 [inlined]
 [2] ntuple
   @ ./ntuple.jl:50 [inlined]
 [3] convert(#unused#::Type{Tuple{Vector{Int64}, Vector{Int64}, Vector{Float64}}}, x::Tuple{Vector{Int64}, Vector{Int64}, Nothing})
   @ Base ./essentials.jl:419
 [4] setindex!(h::Dict{Tuple{Symbol, Symbol, Symbol}, Tuple{Vector{Int64}, Vector{Int64}, Vector{Float64}}}, v0::Tuple{Vector{Int64}, Vector{Int64}, Nothing}, key::Tuple{Symbol, Symbol, Symbol})
   @ Base ./dict.jl:369
 [5] add_edges(g::GNNHeteroGraph{Tuple{Vector{Int64}, Vector{Int64}, Vector{Float64}}}, edge_t::Tuple{Symbol, Symbol, Symbol}, snew::Vector{Int64}, tnew::Vector{Int64}; edata::Nothing, num_nodes::Dict{Symbol, Int64})
   @ GraphNeuralNetworks.GNNGraphs ~/fluxml/GraphNeuralNetworks.jl/src/GNNGraphs/transform.jl:223
 [6] add_edges(g::GNNHeteroGraph{Tuple{Vector{Int64}, Vector{Int64}, Vector{Float64}}}, edge_t::Tuple{Symbol, Symbol, Symbol}, snew::Vector{Int64}, tnew::Vector{Int64})
   @ GraphNeuralNetworks.GNNGraphs ~/fluxml/GraphNeuralNetworks.jl/src/GNNGraphs/transform.jl:172
 [7] add_edges(g::GNNHeteroGraph{Tuple{Vector{Int64}, Vector{Int64}, Vector{Float64}}}, data::Pair{Tuple{Symbol, Symbol, Symbol}, Tuple{Vector{Int64}, Vector{Int64}}}; kws::Base.Pairs{Symbol, Union{}, Tuple{}, NamedTuple{(), Tuple{}}})
   @ GraphNeuralNetworks.GNNGraphs ~/fluxml/GraphNeuralNetworks.jl/src/GNNGraphs/transform.jl:170
 [8] add_edges(g::GNNHeteroGraph{Tuple{Vector{Int64}, Vector{Int64}, Vector{Float64}}}, data::Pair{Tuple{Symbol, Symbol, Symbol}, Tuple{Vector{Int64}, Vector{Int64}}})
   @ GraphNeuralNetworks.GNNGraphs ~/fluxml/GraphNeuralNetworks.jl/src/GNNGraphs/transform.jl:170
 [9] top-level scope
   @ REPL[83]:1

But ofcourse, I can add new edges if the original graph doesn't contain any weights:

julia> g_het2 = GNNHeteroGraph((:user, :rate, :movie) => ([1,1,2,3], [7,13,5,7]))
GNNHeteroGraph:
  num_nodes: Dict(:movie => 13, :user => 3)
  num_edges: Dict((:user, :rate, :movie) => 4)

julia> g_het2 = add_edges(g_het2, (:user, :like, :actor) => ([1,2], [3,4]))
GNNHeteroGraph:
  num_nodes: Dict(:actor => 4, :movie => 13, :user => 3)
  num_edges: Dict((:user, :like, :actor) => 2, (:user, :rate, :movie) => 4)

If this is expected behaviour, it'll be really nice to add these to the documentation.

CarloLucibello / GraphNeuralNetworks.jl

Edge weights not properly documented for `GNNHeteroGraph`s (and implement new function to add new edge weights?) #331