[AutoTVM] Catch cuda errors when using LocalRunner

[michalpiszczek]

This block in measure_methods.py intended to capture CUDA errors caused by invalid kernels generated during AutoTuning:

    except TVMError as exc:
        msg = str(exc)
        if "Stack trace returned" in msg:
            msg = msg[: msg.index("Stack trace returned")]
        if "CUDA Source" in msg:
            msg = msg[: msg.index("CUDA Source")]
        costs = (RuntimeError(msg[:1024]),)
        errno = MeasureErrorNo.RUNTIME_DEVICE

but when using LocalRunner the cuda errors come through here wrapped in RPCError not TVMError.