OPENDJ-2328: Optimize UUID equality matching rule indexing

Made the following changes:

* removed duplicate normalization code between the ordering and equality
matching rules

* changed normalization algorithm to normalize values as 16 byte values
rather than 36 byte strings

* changed equality matching rule indexer to hash normalized values down
to 4 bytes and changed the ordering matching rule so that it no longer
reuses the equality matching rule index (this is ok because UUID
attributes are not normally indexed for ordering).

Benefits:

* 6% (30MB) reduction in disk/cache space for 1M entries
* small but significant boost in performance.

Key collisions:

* 1M entries had 101 key collisions (0.01%)
* 10M entries had 10000 key collisions (0.1%) of which only 7 mapped to
3 entries, the rest only mapped to 2.