Efficiently handle circuit sizes 10x or more of today's circuits

The field of quantum computing is progressing from quantum hardware having small numbers of noisy, short-lived physical qubits, to hardware having 100s–1000s of physical qubits combined algorithmically into error-corrected, long-lived logical qubits. For TensorFlow Quantum to support QML in the early fault-tolerant quantum computing era, it should be possible to handle circuit sizes that are at least 10x what it can comfortably handle today.