Optimize centrality reports using in-memory mutate

[JohT]

🎉 Reports

• Add centrality summary containing all values: After each projection, all calculated centrality values are written to a new report file to be able to compare them and have them all in one CSV.

🚀 Features

• Support empty=no relationship weight parameter: To be able to run Centrality algorithms with and without a relationship weight property, the query parameter is converted to null when containing an empty string. Since only string query parameters are supported for now, null can't be passed directly.
• Add centrality for methods: Now, method invocations are also examined by the Centrality algorithms.

⚙️ Optimization

• Optimize centrality reports using in-memory mutate: Instead of running every algorithm twice (stream & write) they now run only once (mutate). The results are then streamed or written from the in memory projection without recalculating them.
• Raise transaction memory to 3 GB: Fixes problems when preparing large graphs for analysis.
• Write Modularity only for Artifacts & Packages: Writing the Modularity metric values to each node still takes lots of resources in large graphs. Therefore it is only done for artifact and package nodes and only for Leiden communities.
• Rename "without-empty" to "cleaned": This focusses more on the "what" instead of the "how" and is more general for different use cases.

📦 Updates

• Update open graph data science to v2.5.0

🛠 Fixes

• Fix command line input error for explore mode: Starting an analysis in --explore mode didn't do anything. The shift command, that is used to read the next command line argument, fails when there are no more arguments left. Since activating fail-fast the program stopped right after parsing the command line arguments. Ignoring that specific error fixes this problem.