Add memory estimation function

[mfherbst]

trafficstars

It was brought up in psi4/psi4#1744 that a memory estimate for ADC calculations could be a very helpful feature. What could be helpful is to allow the estimation of the memory needed for a Tensor before allocating it, based on its symmetry and generalise the notion from there to the whole of adcc.

[mfherbst]

I'm working on that and have an estimate for the lower memory bound of restricted ADC calculations. The idea is to use this lower bound to check the user does not request something impossible with respect to the available main memory.