Fix runtestmanager for CICD deployments

In CICD deployments, there are no workarounds that install
docker.io before using 'runtest' formula.
Add preparation of necessary packages to runtestmanager.

- Refactor RuntestManager to use salt CLI instead of salt API
  because salt API dones't allow to catch errors from modules
- Add preparations for docker packages, Neutron private and
  public networks based on underlay.yaml address pools, etc.

Change-Id: Iaff8046b9cde6185cd6718a5c2064998ea7dc507
diff --git a/jobs/pipelines/swarm-testrail-report.groovy b/jobs/pipelines/swarm-testrail-report.groovy
index 333547a..e7ffe92 100644
--- a/jobs/pipelines/swarm-testrail-report.groovy
+++ b/jobs/pipelines/swarm-testrail-report.groovy
@@ -31,6 +31,13 @@
     }
     dir("${PARENT_WORKSPACE}") {
         try {
+
+            if (env.TCP_QA_REFS) {
+                stage("Update working dir to patch ${TCP_QA_REFS}") {
+                    shared.update_working_dir()
+                }
+            }
+
             def report_name = ''
             def testSuiteName = ''
             def methodname = ''
@@ -111,7 +118,7 @@
             }
 
         } catch (e) {
-            common.printMsg("Job is failed: " + e.message, "red")
+            common.printMsg("Job is failed", "red")
             throw e
         } finally {
             // reporting is failed for some reason