
The LRU options seems to hurting compression more than it should.

On 'paper1' (51k bytes) compression is 2.285 (coder 0)
with -L70000 compression is 2.300

this seems ludicrous

possible wierdities:

	1. a bug

	2. shared LRU between all orders causing other-order good stuff
			to be flushed out

	3. lots of empty context-adds causing flushing of good stuff
			perhaps switch back to PPMZ_Scale method?
			keep doing scale + updateLRU untill you get one cleared?

the problem is its very hard to really monitor the LRU..

