issue #594, handle SocketTimeoutException differently

[weizhu-us]

It consists of two changes:

• HThriftClient, generates HectorException in getCassandra
• Revert back the change for SocketTimeoutException should be translated as HectorTransportException #434. I have tested the different scenario by shutting down Cassandra or make it responds really slow. I can't observe the problem described in SocketTimeoutException should be translated as HectorTransportException #434. By coverting SocketTimeoutException to HTimedOutException, it avoids shutting down the pool when socket timed out.

The code has been in our production for over 2 weeks and everything works as expected.

[zznate]

This looks good - thanks!