Handle failed flex job template updates

[andremarianiello]

This small changes improves the handling of an updating flex template job by waiting for the new job to be in a RUNNING state. Previously, we were waiting for the old job to terminate in some fashion, which won't happen if the new job fails to launch for some reason. This change ensures that the provider is not stuck in a "Still waiting..." loop until timing out. This issue was mentioned in hashicorp/terraform-provider-google#8408 (comment) on hashicorp/terraform-provider-google#8408

[andremarianiello]

@ScottSuarez please let me know if there is anything more I should do for this.

[ScottSuarez]

Hey sorry, I'm off on vacation until next Tuesday. I'll take a look then. Thanks for your patience

[ScottSuarez]

I see so instead of waiting for the old job to be updated you are now going to wait for the new job to be running. Let me run this test locally since it's skipping the vcr

[ScottSuarez]

@andremarianiello could you make this change upstream to our code generation tool that we use to create tpg and tpgb?

https://github.com/GoogleCloudPlatform/magic-modules/blob/master/mmv1/third_party/terraform/resources/resource_dataflow_flex_template_job.go.erb
https://github.com/GoogleCloudPlatform/magic-modules/blob/master/mmv1/third_party/terraform/tests/resource_dataflow_flex_template_job_test.go.erb

[andremarianiello]

@ScottSuarez GoogleCloudPlatform/magic-modules#4847

[melinath]

Superseded by the upstream PR.