RFR: JDK-8014838: getStrongSecureRandom() should require at least one implementation

[Bradford Wetmore]

This minor suggestion came up in May from our JCK team:

     https://bugs.openjdk.java.net/browse/JDK-8014838

     http://cr.openjdk.java.net/~wetmore/8014838/webrev.00/

JCK suggested all implementations should have at least one strong random 
implementation.

I've also changed the error case to throw NoSuchAlgorithmException 
instead of returning null.

CCC review is also underway concurrently, but I'm not expecting any issues.

Brad