Handle loss in stim control flow

[joao-boechat]

In a REQUIRE statement inside of a PREPARE block, we need to make sure that the targets being required aren't LOSS. If they are, then we have to restart the prepare block instead of just continuing (which is what we currently do).

So for example:

PREPARE {
    LOSS_ERROR(0.2) 0
    M 0
    REQUIRE rec[-1]
}

When plotting the histogram for this, we should only see a 0 result, as the loss leads to a retry.

Meanwhile,

PREPARE {
    LOSS_ERROR(1) 0
    M 0
    REQUIRE rec[-1]
}

should result in an infinite loop as we are always getting loss