Tenants usually read Applications but not the HelmReleases behind them,
and only Ready and Released were copied over. A client could therefore
see that a release was failing but not how often it had failed, nor
whether helm-controller had stopped retrying it.

Copy the Stalled condition, which helm-controller sets once it stops
reconciling a release (remediation retries exhausted, rollback target
missing, invalid CEL, access denied), and surface the install and
upgrade failure counters with the last attempted release action.

Stalled is dropped when it was observed for an older generation than
the HelmRelease's current one: the Application carries no generation,
so after a spec update a client would otherwise read the previous
spec's failure as terminal for the new values.

With the RetryOnFailure policy the platform generates today, a failed
release is retried rather than stalled, so the counters are the signal.
That does not hold for every state: a release missing from storage
while an install failure count is retained still stalls, and a release
on a remediation policy stalls once its retries run out.

Assisted-by: LLM
Signed-off-by: Alexey Artamonov <[email protected]>
(cherry picked from commit 6e36385)