3.1 was released

[tsaloranta]

Quick note: after doing some performance analysis, I found a few minor
things that could be done further improve performance of UUID creation
when using time/location based variant (which is already the most
performant one).
This gets throughput quite close to theoretical maximum that clock
resolution allows (10 million UUIDs per second), depending on system.

For 3.2.0 I hope to figure out a way to do more granular locking: the
remaining challenges are:

(a) Currently all time/location based creation is synchronized; but
ideally it would only be synchronized on per-address (ethernet MAC)
basis -- this would allow creation of more than 10 m/sec, when using
multiple NIC addresses (I admit that this is of more theoretical
interest at this point)
(b) For simple single-JVM, single-threaded access use case, it would
be nice to allow for no-sync use case. This is not trivial to make in
a way that allows most of the code to be shared; but should be doable
with some thinking & clever design :)

-+ Tatu +-