Clients should record workloads they fail to build, and communicate this to the server.

[AndyGrant]

Sometimes a worker fails to build a test. This can occur due to a fault on the part of the test writer; a fault in the engine support mechanism in OpenBench; or possibly other means.

The result is a worker will often continually attempt and fail to build the same test, throwing errors to the client console and to the framework errors endpoints, and also wasting CPU time.

The remedy: Clients should keep a list of test ids that they failed to build, in config.blacklist. This black list can be sent alongside the machine_id and secret in the clientGetWorkload endpoint. As a "super" preliminary step, the server should filter out matching ids.

1. In safe_download_engine(), when catching OpenBenchBuildFailedException, add the test id to the config's blacklist.
2. Pass the configs blacklist in server_request_workload().
3. Pass the request object to get_workload in views.py's client_get_workload
4. Pass this through select_workload and filter_valid_workloads
5. In the very first step here, filter anything in request.POST['blacklist']