K8sJobExecutor: Respect executor settings, specified on job launch

[msavtchouk-pf]

Summary & Motivation

Allow k8s_job_executor configuration on job launch, including step_k8s_config and per_step_k8s_config.
Previously, when run_configuration was specified on the job launch, i.e. using grapql interface, this configuration was completely ignored. See example below.

client = dagster_graphql.DagsterGraphQLClient()

run_config = dag.RunConfig(
    execution={
        "config": {
            "step_k8s_config": {"container_config": {"resources": {"requests": {"cpu": "100m"}}}},
            "per_step_k8s_config": {
                "dyn_sink": {"container_config": {"resources": {"requests": {"cpu": "100m"}}}},
            },
        }
    },
)
client.submit_job_execution(
        dagster_job_name, run_config=run_config)

How I Tested These Changes

I have implemented a test that tests this change. In addition, tested with a patched custom executor on local installation of k8s dagster

Changelog

Insert changelog entry or delete this section.

[gibsondan]

Hi @msavtchouk-pf - thank you for sending this out and I agree the underlying problem here is annoying that you need to set run and step config differently, but this isn't something that we're currently planning to change. executor config currently only affects the executor rather than the run launcher, and its not uncommon for steps and runs to have different configuration (especially resources).

Typically the inheritance works in the opposite direction - configuration on runs is applied down to steps if you're using the k8srunlauncher, but can be overridden at the step level using executor config.

I could imagine us adding some kind of 'run launcher config' to run config akin to executor config that allows you to configure the run launch params as well via the launchpad, but that would be a larger change.