Refactor runners

[alecandido]

Following to the newly chosen benchmark layout the runners should be refactored, backed by a proper library (in banana) and the new runners (in yadism/benchmarks) making use of the library.

• develop runners tools, to support one of the models of Generate OCards in the runners #3
  • they should support parallelization, since the only critical operation it's writing on db (see note on DBMS #2)
• create new runners
  • making use of the tools
  • properly tagged

[alecandido]

At the level of runners it might still be a nice idea to have tags (then we'll propagate to the db and the navigator, as a second step). Something like:

self.run(..., tags=["scale variations", pto, fns]