should not allow submitting results to different servers

[Antonio Terceiro]

Currently the submit_results action allows one to submit results to random servers, but that's not actually feasible since you cannot control which version of the bundle schema the other server runs, and there is no useful use case for submitting results for a server other than then one you are submitting the job to.

So we need to deprecate the "server" argument from the submit_results action, probably making it optional in the schema and ignoring its value. The scheduler probably also handles this on lava_scheduler_app/models.py, so we need to kill dead code there when we drop support for "server" in the dispatcher.

[Antonio Terceiro (terceiro)]

i.e. currently we say:

    {
        "command": "submit_results",
        "parameters": {
            "stream": "/anonymous/terceiro/",
            "server": "http://localhost/RPC2/"
        }
    }

but you should be able to say just

    {
        "command": "submit_results",
        "parameters": {
            "stream": "/anonymous/terceiro/",
        }
    }

and moreover ignore the "server" argument if it exists. We cannot remove it from the schema for backwards compatibility.

[Antonio Terceiro (terceiro)]

we should go even further:

- just submit *all* results without the need to an explicit bundle stream
- add a bundle_stream attribute to the top-level of the JSON job file
- make sure the installation creates a default bundle stream, to be used in the case the bundle_stream parameter is omitted
- still support the explicit submit_results action for backwards compatibility

[Dave Pigott (dpigott)]

Not really high - not affecting anybody - bad config, re-configure