`Pattern.get_layers` assumes that inputs cannot be outputs

[matulni]

Pattern.get_layers assumes that inputs cannot be outputs.

This results in inconsistent behaviour, e.g.,

from graphix.pattern import Pattern

p1 = Pattern(input_nodes=[0, 1], cmds=[E((0, 1))])
print(p1) # E(0,1)

meas_order_1 = p1._measurement_order_depth()
print(meas_order_1) # [0, 1] -> Incorrect, [0, 1] are not measured.

p2 = Pattern(input_nodes=[], cmds=[N(0), N(1), E((0, 1))])
print(p2) # E(0,1) N(1) N(0)

meas_order_2 = p2._measurement_order_depth()
print(meas_order_2) # [] -> Correct

assert set(meas_order_1) == set(meas_order_2) # Error

To be fixed in Flow refactor part 2