Add graph data science algorithms for Typescript

[JohT]

🚀 Reports

• Add graph data science algorithms for Typescript. Typescript module dependencies are now also analyzed by all previously for Java Packages available Graph Data Science procedures. This could be done easily utilizing an already existing abstraction with parametrized node labels for DEPENDS_ON relationship Graph projections.

• Visualize HashGNN and node2vec node embeddings. Additionally to the visualization of Fast Random Projection Node Embeddings, HashGNN and node2vec node embeddings are now also included in the Jupyter Notebook NodeEmbeddings.ipynb so that their results can be compared to each other.

⚙️ Optimization

• Remove lower case workaround after upstream fix.
• Centralize logging for missing projection data. This refactoring leads to a much cleaner creation of Graph projections for dependencies that also include the case where there is no data and a message needs to be logged in centralized, shared functions.
• Use already calculated node embeddings if available. If Node Embeddings have already been calculated by the CSV report NodeEmbeddingsCsv.sh, they shouldn't be calculated again for the Jupyter notebook NodeEmbeddings.ipynb. Otherwise they should to not introduce a sequential dependency that would make it harder to parallelize those reports in the future (scaleability).
• Use undirected Graph projections for all node embeddings. Even though the node2vec implementation of the Neo4j Graph Data Science Plugin does support directed Graph projections, the visualized node embeddings seem to have a clearer separation of communities in the simple examples here. Trying out different hyper parameter and comparing there impact on the created node embeddings could be an idea for a future feature.

🛠 Fixes

• Fix typo in comment
• Add fine grained weights for internal Typescript Module dependencies. This fix is needed for larger Typescript code bases that have inter module dependencies that are directly detected as relationships between "Module" nodes (instead of Module->ExternalModule relationships). This assures that (Module)-[DEPENDS_ON]->(Module) stays homogeneous which is essential for many Graph algorithms.