Simplify hash_table somewhat.

- Use djb2 hash http://www.cse.yorku.ca/~oz/hash.html
  Performs similarly to the existing function, but removes the need for a prime
  number of buckets. Doesn't need to do an strlen every insert either.
- Add some more heuristics. Collected in realtime (cheap), no postprocessing.

git-svn-id: http://opkg.googlecode.com/svn/trunk@336 e8e0d7a0-c8d9-11dd-a880-a1081c7ac358