Create a results database so we can track test outcomes

[hcoyote]

Nice to have.

We do OMB tests often enough with various cluster geometries and workload shapes/sizes, that we probably should start recording tests in a way that we can compare behaviors over time.

Results should probably record:

1. Driver config
2. Workload shape
3. Client count
4. Cluster shape and configuration
5. Client and cluster instance types

Probably also want to record if the run ran to successful completion, and if it did not, any sort of detail about the failure.