Capture "job graph is too large" and provide suggestions

[kellen]

When users experience "job graph too large" issues, the problem is typically implicitly-derived coders or parquet config. We could potentially capture this exception and analyze the graph to automatically suggest a resolution. Alternatively, we can provide a standard set of potential resolutions in an improved error message.

Exception in thread "main" java.lang.RuntimeException: Failed to create a workflow job: (3d8bf2a262152664): The job graph is too large. Please try again with a smaller job graph, or split your job into two or more smaller jobs.
	at org.apache.beam.runners.dataflow.DataflowRunner.run(DataflowRunner.java:1363)
	at org.apache.beam.runners.dataflow.DataflowRunner.run(DataflowRunner.java:198)
	at org.apache.beam.sdk.Pipeline.run(Pipeline.java:321)
	at org.apache.beam.sdk.Pipeline.run(Pipeline.java:307)
	at com.spotify.scio.ScioContext.execute(ScioContext.scala:586)
	at com.spotify.scio.ScioContext$$anonfun$run$1.apply(ScioContext.scala:573)
	at com.spotify.scio.ScioContext$$anonfun$run$1.apply(ScioContext.scala:561)
	at com.spotify.scio.ScioContext.requireNotClosed(ScioContext.scala:652)
	at com.spotify.scio.ScioContext.run(ScioContext.scala:561)
	at com.spotify.data.counters.package$DataCountersPublisherContext.runAndPublishMetrics(package.scala:105)
	at com.spotify.data.counters.package$DataCountersPublisherContext.runAndPublishMetrics(package.scala:110)
	at com.spotify.audience.promo.jobs.PromsToAdPlatformMappingsGCS$.main(PromsToAdPlatformMappingsGCS.scala:75)
	at com.spotify.audience.promo.jobs.PromsToAdPlatformMappingsGCS.main(PromsToAdPlatformMappingsGCS.scala)
Caused by: com.google.api.client.googleapis.json.GoogleJsonResponseException: 400 Bad Request"