Giant tiny bug found in debinning.py for invV_xy. The round-off error "correction" was effectively shrinking the covariance by n^4. Now fixed. I should try all covariant chiSq again.

Also, I changed to a different normalization scheme. Now the debinning function
is no longer normalized to 1, but instead to nData. Probably need to fix up all
examples after that change. Thus, this may be a broken/partial commit.

Changes to be committed:
modified: debinningCore/debinning.py
modified: examples/discriminationMC.py