Enabling pre-submit by addressing the comments received from DevOps team - #119

The new pull-request script is not on main yet, so the PR-119 trial has to load it from feature/CSTACKEX-340. Change this back to main before the folder builds every pull request.

Jenkins warns that approving RunWrapper.getRawBuild lets a pipeline step outside the script sandbox. The existing job lookup already used for older builds can read the same revision.

A manual build does not copy the scan revision onto the build, so the prepare stage treated a real PR as unindexed.

The multibranch job never attaches a pull request revision, so refusing the build there skipped the GitHub eligibility check.

The environment block declared RESOLVED_SOURCE_ID and RESOLVED_SOURCE_SHA
from build parameters. A multibranch pull request has no such parameters,
and the block wraps every stage, so it shadowed the values set in
prepareMultibranchPullRequest(). They read back empty, which caused a
NullPointerException in presubmitSubjectBase and an empty SHA in the
Maven and deploy shell steps. Set them in resolveSourceIdentity() instead.

Also name the trusted CI branch in one place, trustedCiBranch(), and
describe the multibranch folder configuration for the pilot.

Build 8 of PR-119 failed on the upstream, timing-sensitive
ConfigKeyScheduledExecutionWrapperTest, which counts scheduled ticks in a
fixed window and jitters in the CPU-limited agent pod. Add an opt-in
RERUN_FAILING_TESTS setting to mvn-full.sh that passes
-Dsurefire.rerunFailingTestsCount, and set it to 2 in the Jenkinsfile. The
default stays 0 for local runs.

…out.

The multibranch CI checkout fetched only private-cicd, so Phase 2 stopped with
"Trusted ONTAP prerequisite checker is missing" right after the snapshot
revert. run-phase2-remote.sh runs the trusted copy of
test/integration/plugins/ontap/check_ontap_prereqs.py from the CI checkout, so
add that single file to the sparse paths.

The HTML log link of a failed ONTAP test opened the suite stdout log, which
only holds the pass/fail summary, so the reason for the failure was not
visible. Parse the ERROR and FAIL blocks of the suite's results.txt, put them
in a "Failure details" section above the suite log, and link a failed test
straight to its block. Add an Error column to the mail table with a one-line
reason, leading with the CloudStack job error text when there is one, and link
the archived management and agent logs from the failure section.

test_06_start_vm_lun_remapped failed once with "Unable to apply dhcp entry on
router": the guest network's router had reported failed health checks about a
minute earlier, and the test started the VM with no check and no retry.

Add to OntapTestBase a bounded router readiness gate, a deploy wrapper that
waits before and after the deploy (the first deploy creates the router), and a
start wrapper that retries a VM that rolled back to Stopped, at most three
attempts, and raises every attempt's error if they all fail. A single wait
timeout makes the class stop waiting, so a permanently flagged router cannot
slow every VM operation. Use the wrappers in the iSCSI and NFS3 VM tests.

The stage table, the report's Started and Finished rows and the start mail
showed UTC. Display them in IST (UTC+5:30). The stored timestamps and the
stage durations stay UTC, so the arithmetic is unchanged.

When ONTAP rejects a createVMSnapshot that runs while a revert's clones exist,
CloudStack keeps the snapshot record in Error. The retry in test_23 then
succeeds, but the record has no id the test saw, so it was never deleted. Any
Instance Snapshot blocks detach, resize and destroy, so test_24, test_25 and
test_27 failed. Sweep Error-state VM snapshots of both test VMs in tearDown.

trustedCiBranch() named feature/CSTACKEX-340 for the PR-119 pilot. Return main
so the presubmit uses the merged CI scripts for every pull request, and update
the job guide to match. The Jenkins folder still has to switch its Remote
Jenkinsfile Provider branch to main and its filter to PR-* after this merges.

- Complete the Check after the report and archive exist, so its summary names
  the failed stage and links the Jenkins build and the HTML report.
- Retry the completing PATCH three times; a Check left in progress blocks a
  pull request that requires it.
- Publish an already completed Check when the one created at the start is
  missing, and complete a started Check as cancelled when the build is not
  built, so a result always reaches the pull request.
- Document what each build result does to a required Check.

An ontap-coverage profile in the plugin pom, activated by -Dontap.coverage,
attaches JaCoCo to the build that already runs the plugin's unit tests, so no
second test run is needed. mvn-full.sh turns it on with ONTAP_COVERAGE=true and
keeps the report. ontap-coverage.py summarises the unit and functional reports,
with the combined figure computed per source line, and the report generator
embeds the summary in the mail and report.html.

The plugin runs in the management server, so the JaCoCo agent is attached to
its JVM before the first start. marvin-run.sh dumps the data after the iSCSI
and the NFS3 suites. coverage-report.sh merges the dumps and builds the report
against the plugin jar in the deployed management package. Every step degrades
on its own and never fails the integration run.

The coverage summary is built before the report so the mail carries it, then
added to the Check summary and posted as a single comment that later runs
update in place. PRESUBMIT_COVERAGE=false turns the feature off and
PRESUBMIT_COVERAGE_COMMENT=false keeps only the comment off. The comment step
fails on its own if the App lacks write access, without changing the result.