[AMBARI-22805] Blueprints do not handle some failures properly

[adoroszlai]

What changes were proposed in this pull request?

Introduce ABORTED state for topology host requests (cluster creation, scale hosts) to handle errors that happen in the pre-stage phase (during configure cluster task and other topology tasks). Return such state and cause of failure in requests API (cause is returned in request_context).

https://issues.apache.org/jira/browse/AMBARI-22805

How was this patch tested?

Tested the following scenarios:

• cluster installation from scratch:
  • successful one
  • with wrong Kerberos admin credential
  • invalid blueprint (no NAMENODE)
• upscale cluster:
  • directly after install
  • scale request before restart, agent started after server restart
  • both scale request and agent start issued after restart
• upscale cluster with missing Kerberos credentials (same scenarios)
• upgrade from Ambari 2.6.1
• upscale cluster after upgrade

Unit tests:

...
Tests run: 4943, Failures: 0, Errors: 0, Skipped: 34
...
Total run:1201
Total errors:0
Total failures:0
OK
...
BUILD SUCCESS

Sample failed cluster creation:

{
  "href" : "http://localhost:8080/api/v1/clusters/TEST/requests/1",
  "Requests" : {
    "cluster_name" : "TEST",
    "id" : 1,
    "progress_percent" : -1.0,
    "request_context" : "Logical Request: Provision Cluster 'TEST'\nFAILED: Missing KDC administrator credentials.",
    "request_status" : "ABORTED"
  }
}

@rnettleton @mhmxs

[asfgit]

Refer to this link for build results (access rights to CI server needed):
https://builds.apache.org/job/Ambari-Github-PullRequest-Builder/190/
Test PASSed.