fixes bug in merge tablet reservations (#4604)

The fate step that reserves tablets for merge was signaling it
was ready when it was not in the case where its scan of the
metadata table took 0ms.  This was causing some ITs to fail.  The
bug was introduced in #4574

server/manager/src/main/java/org/apache/accumulo/manager/tableOps/merge/ReserveTablets.java

@@ -115,14 +115,10 @@ public long isReady(FateId fateId, Manager env) throws Exception {
     }
 
     long sleepTime = Math.min(Math.max(1000, count), maxSleepTime);
-    if (locations > 0 || otherOps > 0 || wals > 0) {
-      // need to wait on these tablets
-      return sleepTime;
-    }
-
-    if (opsSet != opsAccepted.get()) {
-      // not all operation ids were set
-      return sleepTime;
+    if (locations > 0 || otherOps > 0 || wals > 0 || opsSet != opsAccepted.get()) {
+      // need to wait on these tablets, must return non-zero to indicate not ready so need to handle
+      // case of sleepTime being zero
+      return Math.max(1, sleepTime);
     }
 
     // operations ids were set on all tablets and no tablets have locations, so ready