[Costs] Add a cost metric for T / Toffoli Depth

[tanujkhattar]

If you have enough magic state factories to operate at reaction limited regime, the reaction depth / depth of non-clifford operations becomes the limiting factor for speed of computation. As cost of producing magic states becomes cheaper, the depth of T / Toffoli gates becomes a more relevant cost metric than number of T / Toffoli gates.

This issue is a feature request to add a new CostKey that keeps track of the depth of T / Toffoli gates in the algorithm.