Refactor benchmarks with APFEL

[alecandido]

• reorganize input files (runcards)
  • run from db (with queries)
• reorganize apfel cache
  • hashing input dicts (try with frozendict, or better just hash(str(d)))
• reorganize logs

in order to do that:

• setup a lightweight proper db
• interface with old
  • produce runcards from db
  • produce theory/observables from matrix
  • compare different logs (subtract_tables.py)
  • print comparisons (print_comparison_tables or better)
  • add subtract_tables to ParentTest

finally:

• update the workflow to generate input.json
• download needed PDFSets with lhapdf

[alecandido]

to think about:

• hash vs db ids

to do:

• uniquely identify with theory and observables:
  • logs
  • apfel_cache
• subtract tables: either identify with theory + dis_observables hashes (for logs and apfel_cache) or by the unique id (for scratch area)
  • use it as 's11' or 'scratch11' or something like that to distinguish from hashes
• navigator: list entries in a quick form, output content