Dear all,

we need to check that expressions in CASE selectors are scalar,
and reject them early if they aren't.

Testcase by Gerhard.  The fix is obvious and sort of a follow-up
to the fix for PR103591.

Regtested on x86_64-pc-linux-gnu.

Will commit as obvious within 48h unless there are objections or
other comments.

Thanks,
Harald