Improve Parallel Execution of Clustering in CWL Workflow

The problem is that CWL takes very long to evaluate JS to read the components and files processed in a single batch. Two alternatives would fix this issue:

• Store Componentbatch / ClusteringWorkload in Graph Database and read it as a first step of the clustering. Might require model changes
• Only pass the file(s) produced by the GraphComponents step and read the necessary files and components before the clustering step