tune Uid to byte[] packing

Packing Uids into a serialized form is a very common operation and must be as fast as possible.

There are opportunities for performance improvement in packing the Uid into a byte[], then putting that byte[] into the output buffer, rather than packing the Uid into the output buffer directly. This is due to the performance of the output buffer packing methods (syncronization and flush) which is really a separate problem.

Having the ability to get a byte[] representation of a Uid is also useful for improving Xid creation performance, as it can simply be array copied into the new Xid.