bug#40216: 28.0.50; Misinformation in isearch char-fold

[Juri Linkov]

> Ah, I hadn't considered that use case. Do we need an option to
> char-fold-regexp that says 'only apply char-folding to non-ascii
> characters'? That would reduce the size of the regexp considerably.

Currently there are 2 covered use cases:

1. the default is to fold ascii to non-ascii characters;

2. non-nil char-fold-symmetric additionally folds
   non-ascii to ascii characters.

It seems you are proposing a third use case:

3. symmetric-only that can be implemented with a new non-nil option
   char-fold-symmetric-only that will fold only non-ascii characters
   to ascii.

I have doubts how useful this will be.

The current default behavior is useful when the user types
ascii characters on the keyboard with ascii characters only.

The option char-fold-symmetric is useful to match pasted text
both ways ignoring all differences between ascii/non-ascii characters.

But for symmetric-only I can't imagine any useful use case.
For example, when you paste non-ascii characters into the search string,
and want to find corresponding ascii characters.  But why wouldn't you
want to find the other way around: pasting ascii characters
to find non-ascii counterparts?