I suggest to keep it, maybe rephrase a little that the PR author *should* squash at the end of the review in case there are to many iteration commits or "fixups" or so. When done only once at the end, the force-push shouldn't be so bad and the reviewer needs to diff on the diff only once (as @b4n said). So in the end the current rule keeps recommendation but it's not a must. And it depends on PR itself anyway.