[ML] Some logging improvements

- Downgrade flush logging as it's called too often from datafeeds
- Add log info for job opening

Original commit: elastic/x-pack-elasticsearch@a262069f78

plugin/src/main/java/org/elasticsearch/xpack/ml/job/process/autodetect/AutodetectCommunicator.java

@@ -184,7 +184,7 @@ public class AutodetectCommunicator implements Closeable {
         // clients from querying results in the middle of normalization.
         autoDetectResultProcessor.waitUntilRenormalizerIsIdle();
 
-        LOGGER.info("[{}] Flush completed", job.getId());
+        LOGGER.debug("[{}] Flush completed", job.getId());
     }
 
     /**

plugin/src/main/java/org/elasticsearch/xpack/ml/job/process/autodetect/AutodetectProcessManager.java

@@ -204,6 +204,7 @@ public class AutodetectProcessManager extends AbstractComponent {
     }
 
     public void openJob(String jobId, JobTask jobTask, boolean ignoreDowntime, Consumer<Exception> handler) {
+        logger.info("Opening job [{}]", jobId);
         Job job = jobManager.getJobOrThrowIfUnknown(jobId);
         jobProvider.getAutodetectParams(job, params -> {
             // We need to fork, otherwise we restore model state from a network thread (several GET api calls):