Only do address updates for CASE delivery failures.

When the handling of "failed to send message, so update the IP for this peer"
moved out of DeviceCommissioner, we stopped checking whether the session is a
CASE session.  Go back to checking that again.

Fixes project-chip#22326

src/app/OperationalSessionSetup.cpp

@@ -411,7 +411,7 @@ void OperationalSessionSetup::PerformAddressUpdate()
     CHIP_ERROR err = LookupPeerAddress();
     if (err != CHIP_NO_ERROR)
     {
-        ChipLogError(Controller, "PerformAddressUpdate could not perform lookup");
+        ChipLogError(Controller, "PerformAddressUpdate could not perform lookup: %" CHIP_ERROR_FORMAT, err.Format());
         DequeueConnectionCallbacks(err);
         // Do not touch `this` instance anymore; it has been destroyed in DequeueConnectionCallbacks.
         return;

src/messaging/ReliableMessageMgr.cpp

@@ -320,7 +320,8 @@ CHIP_ERROR ReliableMessageMgr::SendFromRetransTable(RetransTableEntry * entry)
         if (exchangeMgr)
         {
             // After the first failure notify session manager to refresh device data
-            if (entry->sendCount == 1 && mSessionUpdateDelegate != nullptr)
+            if (entry->sendCount == 1 && mSessionUpdateDelegate != nullptr && entry->ec->GetSessionHandle()->IsSecureSession() &&
+                entry->ec->GetSessionHandle()->AsSecureSession()->IsCASESession())
             {
                 mSessionUpdateDelegate->UpdatePeerAddress(entry->ec->GetSessionHandle()->GetPeer());
             }