Only expand the propagators when it is applied to the states.

- Remove it from the input parameters, it makes little sense to provide them to the circuit and then do the simulation. If the unitaries are known. it can be trivally multiplied.
- Remove U_list from the properties of CircuitSimulator. It is generated during the process and the dimensions of the unitaries are not determined. It should not be exposed to the the user.

- The method `_process_ops` was executed during the construction, which could be expensive. This is moved to the first execution of the circuit.
- Remove some global members from the class to increase the clearity.

The parameters that will be updated during the circuit simulation are changed to private.

Move all the calculation related to unitary-state multiplication to _evolve_state.

Provide qeye as the input states.

This option computes the full unitary of all the gates, and caches it. It is beneficial if the circuit has a small number of qubits but a very large depth. This is a rare use case.

At the same time, this feature makes the migration to the new efficient circuit evaluation very hard. If the unitary for the circuit is desired, one can just call QubitCircuit.compute_unitary to get the circuit and then apply it to different states.

Use np.einsum to apply the gate matrix directly on the state vector, instead of expanding it first to the full dimension.
- The private member self._state will remain a numpy array during the sequential application of gates.
- If a measurement occurs, the state has to be transformed to a Qobj.
- Returning the state after each step is removed as it inevitably creates Qobj and creates a copy in v4.

…lator_refactor.py