Fix bug in PartitionEdge with mixed vertex types, generalize partitioned_vertices to AbstractGraph

[JoeyT1994]

This is a quick PR to fix a bug and make a few small minor changes:

1. A bug was fixed in the definition of PartitionEdge which was preventing the construction of a PartitionEdge from a NamedEdge{Tuple{V,V'}} where V and V' are different types
2. partitioned_vertices(g; args...) which passes to various partitioning backends now acts on an AbstractGraph type so one can directly pass types such as a DataGraph or ITensorNetwork to it.

[codecov-commenter]

Codecov Report

Attention: 31 lines in your changes are missing coverage. Please review.

Comparison is base (40f0edd) 78.83% compared to head (e9d0402) 77.95%.
Report is 7 commits behind head on main.

Files	Patch %	Lines
...aphs/partitionedgraphs/abstractpartitionedgraph.jl	37.14%	22 Missing ⚠️
src/Graphs/partitionedgraphs/partitionedgraph.jl	76.00%	6 Missing ⚠️
src/Graphs/partitionedgraphs/partitionedge.jl	0.00%	2 Missing ⚠️
.../Graphs/partitionedgraphs/abstractpartitionedge.jl	0.00%	1 Missing ⚠️

❗ Your organization needs to install the Codecov GitHub app to enable full functionality.

Additional details and impacted files

@@            Coverage Diff             @@
##             main      #46      +/-   ##
==========================================
- Coverage   78.83%   77.95%   -0.88%     
==========================================
  Files          31       31              
  Lines        1167     1184      +17     
==========================================
+ Hits          920      923       +3     
- Misses        247      261      +14     

☔ View full report in Codecov by Sentry.
📢 Have feedback on the report? Share it here.

[mtfishman]

I was only suggesting having subgraph(::PartitionedGraph, ::PartitionVertex) return an unpartitioned graph, I still think subgraph(::PartitionedGraph, ::Vector{<:PartitionVertex}) should return a partitioned graph.

[mtfishman]

Some inspiration for that design comes from BlockArrays.jl:

julia> using BlockArrays

julia> a = BlockArray(randn(4, 4), [2, 2], [2, 2])
2×2-blocked 4×4 BlockMatrix{Float64}:
 -0.00134562   0.716063  │  -0.0813348  -0.777815
  0.219591     1.08181   │  -0.0702123   1.43331 
 ────────────────────────┼───────────────────────
 -0.64773     -0.178246  │   0.0260001  -1.36855 
 -0.355566    -0.3518    │  -0.203171    3.71155 

julia> a[Block(1, 1)]
2×2 Matrix{Float64}:
 -0.00134562  0.716063
  0.219591    1.08181

julia> a[BlockRange(1:2, 1:1)]
2×1-blocked 4×2 BlockMatrix{Float64}:
 -0.00134562   0.716063
  0.219591     1.08181 
 ──────────────────────
 -0.64773     -0.178246
 -0.355566    -0.3518  

i.e. indexing with just a single block returns a non-blocked array, while indexing with a range of blocks preserves the block structure.

[JoeyT1994]

I see. I actually think that is already the behavior so I have reverted it back to that. I just forgot to call it that way in ITensorNetwork.jl

[mtfishman]

Oh yeah, I see that defined here: https://github.com/mtfishman/NamedGraphs.jl/blob/v0.1.16/src/Graphs/partitionedgraphs/abstractpartitionedgraph.jl#L168-L176