[Reporting] Do not swallow errors from trying to save completed job data

[tsullivan]

Kibana 6.2.x:

1. Set http.max_content_length in Elasticsearch to something really low, like 1mb
2. Run a CSV report to generate output greater than 1mb, but less than xpack.reporting.csv.maxSizeBytes
3. Expected results:
   • See a log message in the Kibana server with the HTTP (should be a 413 status code) error message that came from Elasticsearch. ES does not need to log the error.
   • The job data for the failed report should have a useful error message, not "Max attempts reached (3)"

Actual result: any useful error message seems to be swallowed - never logged properly

Related: A warning should be logged if the xpack.reporting.csv.maxSizeBytes in Kibana is greater than http.max_content_length in ES: #26383

[elasticmachine]

Pinging @elastic/kibana-app

[joelgriffith]

💥 starting on this

[joelgriffith]

Took some time to get to the bottom of it, but here's a preliminary PR: #28629

[joelgriffith]

Downstream PR is merged (as is the backport). Closing!