Performance benchmarks: Better configuration

### Feature description

Currently, we're using bash scripts to configure a Docker container (in which Conduit will be running) and also to create test pipelines. There are two implications of this:
1. The configuration for a single test is split into two files (run-workload.sh which runs a Docker container, and workload scripts which create pipelines)
2. Workload scripts are pretty verbose (cURL calls)

Once we're done with [pipeline config files](https://github.com/ConduitIO/conduit/issues/32), we can easily configure a whole performance test with a single YML file (one part of it would configure the test instance, another would the pipeline YML config).

Depends on:
* https://github.com/ConduitIO/conduit/issues/32