Add DaggerGraphs subpackage

[jpsamaroo]

This was started during JuliaCon with @pszufe but I never finished it - it's a naive Graphs.jl directed graph implementation that uses Dagger to partition the graph.

@gdalle

Todo:

• [x] Optimize add_vertices! implementation
• [x] Add undirected graph support
• [x] Add freeze for creating immutable DGraph
• [x] Optimize immutable DGraph operations
• [x] Default to directed graph
• [x] Add vertex/edge metadata support
• [x] Add lazy weights implementation (calls weights on metadata)
• [x] Rename "external" to "background" for cross-partition AdjList
• [x] Switch AdjList to dense representation (like SimpleGraph)
• [x] Optimize purely local operations
• [x] Validate Graphs.jl interface compatibility
• [x] Return and cache neighborhood on remote vertex lookup (inneighbors/outneighbors)
• [ ] Implement user-configurable caching strategies
• [ ] Prefetch neighborhoods
• [x] Implement batched and/or streaming edge creation
• [x] Add loading/saving from $k^2$ files
• [x] Validate benchmark performance on purely-local operations versus SimpleGraph
• [ ] Add docs
• [x] Add CI
• [ ] (Optional) Add support for ingesting vertices/edges from a DataFrame or Tables source