Taking into account both T gates and Toffoli gates in the cost estimation

[PabloAMC]

Right now only Toffoli gates are taken into account for the estimation of time and physical qubits required to synthesize non-Clifford gates in the circuit. https://github.com/quantumlib/OpenFermion/blob/cf53c063d0f124a02ff8776bb7f8afb110d4bde6/src/openfermion/resource_estimates/surface_code_compilation/physical_costing.py#L176-L184 From a Toffoli gate, 2 T gates might be obtained (https://quantum-journal.org/papers/q-2019-04-30-135/pdf/). However, the cost is not exactly 1->2 as the conversion takes time etc. I think adding T gates would be a great idea, in part because T gates are more natural when it comes to rotations, even if it seems that all state-of-the-art methods heavily rely on Toffolis.