[BUG] Commissioning reads have a timeout that's too low

[bzbarsky-apple]

Reproduction steps

See https://github.com/project-chip/connectedhomeip/actions/runs/12271203864/job/34238154369

We are failing commissioning because we do a read and time out after 2 seconds, while the server takes 4 seconds to produce a response.

But spec says:

Whenever the Fail-Safe timer is armed, Commissioners and Administrators SHALL NOT consider any cluster operation to have timed-out before waiting at least 30 seconds for a valid response from the cluster server.

This used to work correctly before #36603, but that PR replaced this logic in AutoCommissioner::GetCommandTimeout:

        timeout = app::kExpectedIMProcessingTime;
...
    auto sessionHandle = device->GetSecureSession();
    if (sessionHandle.HasValue())
    {
        timeout = sessionHandle.Value()->ComputeRoundTripTimeout(timeout);
    }

    // Enforce the spec minimal timeout.  Maybe this enforcement should live in
    // the DeviceCommissioner?
    if (timeout < kMinimumCommissioningStepTimeout)
    {
        timeout = kMinimumCommissioningStepTimeout;
    }

    return MakeOptional(timeout);

with:

    const auto timeout = MakeOptional(app::kExpectedIMProcessingTime); // TODO: Save timeout from PerformCommissioningStep?

which was not obviously an issue because our timeout logic is so convoluted.

Bug prevalence

Always, but only affects commissioning sometimes, since this is timing-dependent

GitHub hash of the SDK that was being used

9c8a552

Platform

core

Platform Version(s)

No response

Anything else?

No response