One Bernoulli observation: the size-$0.05$ MP test must randomise at $X=1$

Let X be a Bernoulli r.v. with parameter $\theta$. Suppose we wish to test $H_0:\theta=1/3$ against $H_1:\theta=2/3$, based on random sample of size 1. Then: