Feature Request: gate data property to distinguish measurement gates and heralded errors

[m-mcewen]

stim.gate_data is a very nice way to test against whole classes of gates in a clear way that's robust to future-new-gates:

if stim.gate_data(op.name).is_reset():
    # handle all gates that reset the qubit: R, RX, RY, MR, MRX, MRY

The same concept is hard to apply to various groups of measurement gates because of some quirks of their implementation.
For instance, trying to case out the single qubit M gates [M, MX, MY, MR, MRX, MRY] leads to conflicts with other gates.
HERALDED_ERASE and HERALDED_PAULI_CHANNEL_1 have almost everything in common with the 1Q M gates: they are non-unitary, 1Q gates that produce measurement. M gates support classical noise on their output, so they are 'noisy gates' as well.

{
n for n, gd in stim.gate_data().items()
if (
    gd.is_single_qubit_gate
    and gd.produces_measurements
)}
>>> {'HERALDED_ERASE', 'HERALDED_PAULI_CHANNEL_1',  'M',  'MR', 'MRX', 'MRY', 'MX', 'MY'}

{
n for n, gd in stim.gate_data().items()
if (
    gd.is_single_qubit_gate
    and gd.produces_measurements
    and not gd.is_noisy_gate  # sadly, measurements are also 'noisy'
)}
>>> set()

It would be nice to have some property to split these two.

Simplest for a reader might be is_measurement_gate, where you define it by fiat to be just the 'measurement gates' [M, MZ, MY, MR, MRX, MRY, MZZ, MXX, MYY, MPP], which might be a more commonly wanted category than produces measurement which includes MPAD and the heralded errors as well.

Another option could be adding something like is_quantumly_noisy which gathers the noisy gates other than the measurement ones.