Fix heartbeat self-conflict causes unnecessary lease release after transient timeout

[Arkatufus]

Summary

• When a heartbeat PUT succeeds on the API server but times out on the client, the retry uses a stale version and gets a CAS conflict. If the conflict response shows the same owner, the lease was never actually lost — but the actor was unconditionally releasing it.
• The Left<> handler in the Granted state now checks whether the conflict owner matches the current node. Same owner: update the version and continue heartbeating. Different owner: release the lease (unchanged behavior).
• Affects both Kubernetes and Azure lease actor implementations.

Changes

LeaseActor.cs (both Kubernetes and Azure): Added an owner equality check before the existing unconditional release in the heartbeat conflict handler. Uses string.Equals with StringComparison.Ordinal for strict matching.

LeaseActorSpec.cs (both Kubernetes and Azure): Added 3 reproduction tests per implementation:

• HeartbeatSelfConflictAfterTimeoutShouldStayGranted — timeout → retry → self-conflict → stays granted
• HeartbeatSelfConflictShouldNotCallLeaseLostCallback — lease-lost callback must not fire on self-conflict
• HeartbeatConflictWithDifferentOwnerAfterTimeoutShouldRelease — guards that genuine conflicts still release

Test plan

• All 6 new tests pass (3 Kubernetes + 3 Azure)
• All existing lease actor tests pass on both platforms
• Azure integration tests pass with Docker/Azurite

[Aaronontheweb]

Minor nitpick

[Aaronontheweb]

resource.Value.Owner and _ownerName are always the same so we could clean this log message up and simplify it.

[Aaronontheweb]

LGTM