Optimize UsageJobDaoImpl.updateJobSuccess to use direct UPDATE instead of row-level lock

Replace lockRow + explicit transaction with a single update(jobId,
jobForUpdate) call. All field values are parameters passed by the
caller — nothing is read from the locked row.

The TransactionLegacy.open(USAGE_DB) is kept since this method
explicitly targets the usage database. The txn.start/commit/rollback
are removed since the single update() call is autocommit.

The lock was redundant because the usage server model is single-owner:
one server processes one job at a time.

Author: sliceofapplepie

engine/schema/src/main/java/com/cloud/usage/dao/UsageJobDaoImpl.java

@@ -61,21 +61,15 @@ public long getLastJobSuccessDateMillis() {
     public void updateJobSuccess(Long jobId, long startMillis, long endMillis, long execTime, boolean success) {
         TransactionLegacy txn = TransactionLegacy.open(TransactionLegacy.USAGE_DB);
         try {
-            txn.start();
-
-            UsageJobVO job = lockRow(jobId, Boolean.TRUE);
             UsageJobVO jobForUpdate = createForUpdate();
             jobForUpdate.setStartMillis(startMillis);
             jobForUpdate.setEndMillis(endMillis);
             jobForUpdate.setExecTime(execTime);
             jobForUpdate.setStartDate(new Date(startMillis));
             jobForUpdate.setEndDate(new Date(endMillis));
             jobForUpdate.setSuccess(success);
-            update(job.getId(), jobForUpdate);
-
-            txn.commit();
+            update(jobId, jobForUpdate);
         } catch (Exception ex) {
-            txn.rollback();
             logger.error("error updating job success date", ex);
             throw new CloudRuntimeException(ex.getMessage());
         } finally {