Investigate profiling and optimizing of chunk generation about bravo HOT 2 CLOSED

Boring, watertable, and simplex all take over a second to run, which is not super-surprising since they all touch a lot of blocks. However, it definitely should be possible to optimize those somewhat.

Some interesting things so far:

• https://gist.github.com/726223 inlines block slot lookups to locals. This patch has no measurable effect on performance.
• Simplex lookups are cheap; simplex itself appears to run almost exactly as fast as boring and faster than watertable, and the lookups don't appear to have any cost.
• Additionally, if simplex doesn't need to reseed, only about 50ms are saved per iteration, and simplex only reseeds when it switches to a new World, currently once per server.

from bravo.

Pushed a handful of improvements to beta/chunk.py as well as a tool to benchmark this kind of thing in the future.

from bravo.