Make empty data construction error message less dev-y

[cpcloud]

When one tries to construct a DataFrame where the data column count and the passed in column count differ the error message is completely unintuitive unless you know something about the internal representation of the data inside of NDFrames.

I think we should at the very least make the dimensions in the error message consistent with how the frame looks rather than how it's implemented.

this code:

df = DataFrame(np.empty(0), columns=list('abc'))

raises this exception:

ValueError: Shape of passed values is (1, 0), indices imply (3, 0)

see here for another example: blaze/blaze#466

[jreback]

sure