[CI] Online c_per_u maintenance and query planner?

[Hideaki Kimura]

I'm now refactoring Section 4 (ESTIMATING CT COST)

As Sam commented, there's some fluctuation in the paper about whether 
the c_per_u statistics is maintained at runtime and used by query 
planner for determining whether or not it rewrites query.

I suggest a simpler approach; CT is always used once it was built,
and c_per_u statistics is used only during designing. Assuming this,
I'm going to cut the latter half of Section 4 by just saying "we used
DS for estimating cardinality".

Does it sound right?
-- 
Hideaki Kimura <hkimura at cs.brown.edu>