Bug fix: ISA check inside control flow operations whose body is defined in a separate circuit

[yaelbh]

#1916 fixed a bug related to connectivity inside control operations. @jakelishman has pointed out a case where the fix does not apply correctly. The case is exemplified in the test test_isa_inside_condition_block_body_in_separate_circuit, which is added in the current PR. Running the test, either on Shrebrooke or Cusco, results with

qiskit_ibm_runtime/utils/utils.py:61: in _is_isa_circuit_helper
    qargs = tuple(qubit_map[bit] for bit in instruction.qubits)
_ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _

.0 = <tuple_iterator object at 0x7f3c30d83af0>

>   qargs = tuple(qubit_map[bit] for bit in instruction.qubits)
E   KeyError: Qubit(QuantumRegister(2, 'inner'), 0)

The current PR fixes the error, and makes the test pass.

The fix is copied from qiskit's GatesInBasis pass, from a previous version where the pass was still written in Python.