Evaluation scripts should read metadata either from the job_metadata.json file or directly from the output CIF

[marcuscollins]

Currently we extract metadata about an experiment or generation "trial" from the path to a file called "refined.cif", e.g.,
/data/sampleworks-exp/occ_sweep/grid_search_results_rf3_chiral_off/1VME_0.75occA_0.25occB/rf3_X-RAY_DIFFRACTION/pure_guidance/ens8_gw0.1/refined.cif (on the shared sampleworks instance).

Currently the discovery and parsing of these directories is done by sampleworks.eval.grid_search_eval_utils.scan_grid_search_results which recursively searches a specified directory (like /data/sampleworks-exp/occ_sweep/grid_search_results_rf3_chiral_off/) for target files (by default, "refined.cif").

We'd like to modify this so that the metadata used by scan_grid_search_results to construct the Trial object comes from a cleaner and more extendable source. It should first come from the CIF file itself, using the category name sampleworks (see #209 which adds this into our CIF files). If that category is not available in the CIF file, the next source should be the file job_metadata.json in the same directory as the generated CIF file (again, generally "refined.cif")