Avoid sporadic "No matching job found" warning

We currently get a "No matching job found - maybe another server was a
bit faster. ;)" warning when we just started the last job of a group.
The reason is that the INQ count going to zero is tracked in the
database by the sql triggers but is not yet updated in the in-memory
copy.

Decrement group_jobs_inq after we loaded a job.

mxqd.c

@@ -1407,6 +1407,17 @@ unsigned long start_job(struct mxq_group_list *glist)
     mx_log_info("   job=%s(%d):%lu:%lu :: started. pid=%d",
         group->user_name, group->user_uid, group->group_id, job->job_id, pid);
 
+    /* The group counts in the database were updated by the sql triggers when
+     * we set the job from ASSIGNED to LOADED. We would pick that up in the
+     * next round of the main loop. Update the in-memory counts right now so
+     * that we don't try to start a new job when there are no INQ jobs left.
+     * This avoids a "No matching job found - maybe another server was a bit
+     * faster" warning when we started the last INQ jobs from a group.
+     */
+
+     group->group_jobs_inq--;
+     group->group_jobs_running++;
+
     return 1;
 }