fix(acceptance): run occ upgrade after enabling the app under test

[DeepDiver1975]

Problem

Acceptance jobs (e.g. owncloud/firstrunwizard) started failing with every
@BeforeScenario hook reporting:

TestHelpers\AppConfigHelper::getCapabilities expected HTTP status 200 but got 503
<status>failure</status> <statuscode>503</statuscode> <message>Service unavailable</message>

All scenarios are then skipped ("7 scenarios (7 skipped)") and the job fails
with exit code 2 — even though the suite reports "There were no unexpected
failures".

Root cause

status.php reports "needsDbUpgrade":true. When ownCloud needs a DB upgrade
it answers every request with HTTP 503 until occ upgrade runs.

The harness installs core from the daily-master-qa build, which already
registers a version of the app under test in the database. The Setup app
step then copies the checked-out app over it and runs occ a:e, but never
upgrades. When the checked-out info.xml version is higher than the
registered one, the instance flags needsDbUpgrade and serves 503 everywhere.

This was triggered in firstrunwizard immediately after its info.xml was
bumped 1.3.0 → 1.4.0 while the QA build still shipped 1.3.0. Confirmed by
comparing the last green run (info.xml 1.3.0, needsDbUpgrade:false) against
the failing run (info.xml 1.4.0, needsDbUpgrade:true) on the same infra
~1h apart.

Fix

Add an explicit occ upgrade --no-app-disable step after all apps are enabled,
for both the primary and federated instances. Exit code 3 (ERROR_UP_TO_DATE
— "already up to date") is treated as success via || [ $? -eq 3 ], so
instances that need no upgrade are unaffected (important since the step runs
under bash -e).

Verification

• YAML validated with yaml.safe_load.
• federated-root is always emitted as a step output (empty when federation is
  off), so the if [ -n "${FEDERATED_ROOT}" ] guard skips the federated
  upgrade cleanly for single-server jobs.

🤖 Generated with Claude Code

[phil-davis]

LGTM