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.

OpenFermion/src/openfermion/resource_estimates/surface_code_compilation/physical_costing.py

Lines 176 to 184 in cf53c06

	def cost_estimator(num_logical_qubits,
	num_toffoli,
	physical_error_rate=1.0E-3,
	portion_of_bounding_box=1.):
	"""
	Produce best cost in terms of physical qubits and real run time based on
	number of toffoli, number of logical qubits, and physical error rate.
	"""

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.