[Performance] Optimize operator - operator multiplication

[jpmoutinho]

Currently different size operators in the apply_operator product are handled by padding the smaller one with identities. Another option is to do it by manipulating the qubit indices without padding, which is likely to be more efficient.