create a benchmark based on DrugMechDB

[andrewsu]

DrugMechDB is a database of drug mechanisms expressed as paths through (or subgraphs of) a knowledge graph. More info: https://github.com/SuLab/DrugMechDB

@cbizon and I will take the lead (at least initially) on trying to use this resource to create a benchmark. (It might possibly build on https://github.com/Carolina1396/drug_mechanisms_rare_diseases_BioThingsExplorer, but maybe not...)

[andrewsu]

Next step, I will extend this notebook https://github.com/SuLab/DMDB_Analysis/blob/main/1_code/1_basic_dmdb_analysis.ipynb to output a CSV modeled on this file https://github.com/TranslatorSRI/Benchmarks/blob/main/MostPrescribed_2019/Data/ConditionToTopDrugs.txt, initially focusing specifically on the Drug - Protein - BiologicalProcess - Disease metapath (the second entry on this slide https://docs.google.com/presentation/d/1hF5ceFCBVdImS3-fDaN9SGafjcItWMDU5teSNn3SXs0/edit#slide=id.g113900f19fd_0_335).

[andrewsu]

Created the PR above based on new code in this notebook in SuLab/DMDB_Analysis@d690406. @cbizon is this enough for me to punt this over to you? I'm not entirely sure that I set up benchmark_definition.json and Dr_Pr_BP_Di_mappings.json correctly...