Refactoring

[dufourc1]

Major refactoring to modernize the codebase, add support for decorated graphs (weighted, categorical edges), and improve documentation.

Core Architecture

• New modular structure: Replaced monolithic implementations with SymArray.jl (fast symmetric arrays), GreedySuffStats.jl, and clean API layer
• Preprocessor system: Unified edge type handling (binary, categorical, continuous) via preprocessor/
• Pseudo sufficient statistics framework: Generic abstraction for different edge distributions in pseudo_suff_stats/
• Removed: src/old/, deprecated src/UI.jl, legacy src/assignments/

Extensions

• PythonOptimalTransport: srGW and fngw algorithms for partition alignment
• BootstrapExt, LightMCExt, MakieExt: Optional functionality via weak dependencies

API Changes

# New high-level functions for different edge types
nethist_binary(A, k)      # Binary networks
nethist_categorical(A, k) # Discrete edge values
nethist_continuous(A, k)  # Real-valued edges

# Graphons.jl integration (re-exported)
# Hungarian algorithm for label alignment
# OT-based partition alignment methods

Documentation

• Added tutorials: simple graphs, weighted networks, multiplex networks, temporal networks
• Custom sufficient statistics example
• Removed outdated internals/rules documentation

Infrastructure

• Dependabot configuration
• Benchmark PR workflow
• CondaPkg.toml for Python dependencies