bug#65797: `buffer-match-p` should not use `func-arity`

[Eli Zaretskii <eliz@gnu.org>]

> From: Stefan Monnier <monnier@iro.umontreal.ca>
> Cc: dmitry@gutov.dev,  joseph@breatheoutbreathe.in,  philipk@posteo.net,
>   mattias.engdegard@gmail.com,  65797@debbugs.gnu.org
> Date: Sat, 14 Oct 2023 10:31:29 -0400
> 
> > What bad things will happen if we leave emacs-29 with no changes?
> 
> Nothing too serious.
> `buffer-match-p` is new in Emacs-29 and it is documented (in the Texinfo,
> not in the docstring) to provide a behavior we're unable to implement.
> So the main aim of the patch is to "fix" this new API so it can be
> implemented as documented.
> 
> But the problem is somewhat corner-case, so it's not super urgent to fix it.
> 
> OTOH, the change is minor and fairly safe.
> Basically the patch replaces an &optional with a &rest in the API.
> Besides allowing more cases (which is mostly a non-issue in terms of
> backward compatibility), this introduces just 1 potential problem:
> 
> When `buffer-match-p` is called without the formerly optional arg, it
> will call the predicate functions with one fewer arg than before.
> 
> The Emacs-29.1 code has a hack that tries to detect when the predicate
> expects "one fewer arg" and if so calls it without the optional arg, so
> I just extended that hack to also handle that reverse problem (when
> there is one fewer arg than expected by the function).
> 
> If we put it into `master`, I guess we can stick to the original hack
> and hope the above "just 1 potential problem" won't bite us, but it
> seems we may as well use the more backward compliant option and put it
> into `emacs-29`.
> 
> See below my current "safe" choice.

Sigh.  I guess we can install this on emacs-29 and cross the
fingers...