ENH/API: add numeric_only with a default of False to numerics

[jreback]

for:

min,max,sum,var,std,kurt,median,mad,prod,skew

The big reason to do this is a performance hit when you happen to have an object column. and you don't realize it.

See: http://stackoverflow.com/questions/18701569/pandas-dataframe-mean-very-slow-how-can-i-calculate-means-of-columns-faster

_reduce in Frame already supports this (and Series will ignore it), so should be trvial to implement

alternatively could just pass **kwds thru (I think we had an issue about this earlier)

[jreback]

related? : #4365

[jbrockmendel]

-1. Working in/around DataFrame._reduce and groupby operations that have numeric_only, it is a PITA and hides bugs.

[jreback]

yep we can just deprecate this i think; do we have an issue for that?