Fix circuit.append() to use protocols.commute to verify if two operations can be "crossed" or not.

[tanujkhattar]

Description of the issue
Since #4870 breaks the assumption that two operations acting on disjoint qubits always commute, circuit.append() should use protocols.commute to find the leftmost moment in which an operation can be inserted.

As part of the work done for CCO, this was implemented by adding a special case of checking whether the current operation is a CCO and if yes, whether the previous moment has a measurement with matching control keys.

Cirq/cirq-core/cirq/circuits/circuit.py

Lines 1972 to 1975 in 90679d2

	if moment.operates_on(op_qubits) or (
	op_control_keys & protocols.measurement_key_objs(moment)
	):
	return last_available

This is problematic because it works only in one direction and breaks optimizers which relies on doing operations on the circuit in reverse order.

How to reproduce the issue

>>> import cirq
>>> q =  cirq.LineQubit.range(2)
>>> c = cirq.Circuit(cirq.Moment(cirq.X(q[0])), cirq.Moment(cirq.measure(q[1], key="m")), cirq.Moment(cirq.X(q[0]).with_classical_controls("m")))
>>> print(c)
0: ───X───────X───
              ║
1: ───────M───╫───
          ║   ║
m: ═══════@═══^═══
>>> cirq.AlignLeft().optimize_circuit(c) # Works as expected.
0: ───X───X───
          ║
1: ───M───╫───
      ║   ║
m: ═══@═══^═══
>>> cirq.AlignRight().optimize_circuit(c) # Does not work as expected because implementation depends on inserting operations in the reverse circuit.
>>> print(c) # Wrong output
          ┌──┐
0: ───X────X─────
           ║
1: ────────╫M────
           ║║
m: ════════^@════
          └──┘

Cirq version
0.14.0.dev

[95-martin-orion]

Cirq sync: we support the general idea here (ensure append works in both directions), but we need to be careful about using protocols.commute for it. For example, appending X(q1) after CX(q0, q1) should keep the X after the CX, even though the two commute.