bug#81372: 30.2; make setopt warn on type mismatch in more cases

>> Here's my current suggestion.  It includes FIXME comments in
>> `setopt--set-local` about which I'd like to have Stephane's opinion, tho
>> this shouldn't stop us from moving with the `setopt--set` part.

Pushed to `master`, closing.

> I think the reason this was implemented for set-local only was due to
> file/directory variables being less intentional than someone's
> configuration in which setopt would be under their full control.

Agreed, but I think this suggests it should be moved outside of
`setopt-local` itself so it applies only to the file/dir-local
variable settings.

> We could improve and share type checking code for both setopt and
> setopt-local and rename the custom variable setopt-local-type-mismatch to
> something like setopt-type-mismatch or create two variables one for each
> context or otherwise have it be a cons global . local treatment like (setq
> setopt-type-mismatch (cons 'accept 'discard)).

That's another option.  But I think this kind of control is weird when
`setopt(-local)` is called from ELisp.


=== Stefan