HALON-895: fix cluster stuck after TS-node crash regression

[andriytk]

The regression was introduced at commit 619bc5b0 in this code:

+isRCNode :: NodeId -> Process (Bool)
+isRCNode nid = do
+    self <- getSelfPid
+    whereisRemoteAsync nid labelRecoveryCoordinator
+    void . spawnLocal $ receiveTimeout (1000000) [] >> usend self ()
+    receiveWait
+      [ match (\(WhereIsReply _ mp) -> (if isNothing mp then return False else return True))]
+    where
+      labelRecoveryCoordinator = "mero-halon.RC"

There were two problems about the timeout here: 1) the timeout event from the spawned thread was ignored and 2) the timeout thread was not cancelled in normal (non-timeout) case. So it was always sending the `()' event to itself which could cause all sorts of weird behaviours (including our one).

Now we don't spawn the timeout thread at all, but directly use receiveTimeout instead of receiveWait.

[vvv]

Looks perfect to me.

[vvv]

merged

[rajanikantchirmade]

Looks good to me.

[andriytk]

assigned to @rajanikant.chirmade

[andriytk]

@rajanikant.chirmade @vvv please review it.

[andriytk]

changed the description