Simplify and speed up the test_compress test, add a test run without compression (better testing of no compression than "nop").

Michael, I took a look at this test because it takes forever to run, mostly because we were opening/closing a cursor on every insert, and inserting 10,000 40KB K/V pairs.

Anyway, I'm not just pushing this change into the tree because I removed a whole bunch of magic from the test. I couldn't see any point in it, but maybe I'm missing something.

Can you please take a fast look at the original, particularly the magic at the end of the script, and if you agree it's not necessary, pull my version into the tree?

Thanks.

michaelcahill