bug#66394: 29.1; Make register-read-with-preview more useful

[Alfred M. Szmidt]

Confirmation, we already have bookmarks which are the "longer"
variants.  Making registers confirm by default, makes registers on
par, as slow to use as bookmarks to use.  If those who wish to use
registers with confirmation, why not use bookmarks?

What becomes the difference between registers and bookmarks with this
change (confirmation part) other than bookmarks being saved to a file?

Some people seemed to have lots and lots (dozens if not more) of
registers, what heppens if you restart Emacs?  Do you save registers
between invocations of Emacs?  In that case, it sounds like one should
be using bookmarks and not registers.

TLDR: Making the default to "confirm" (preview is not the main
problem) makes registers just like bookmarks.  

   here the last patch that provide Emacs-29 previous behavior as asked on 
emacs-dev,
   I made it the default for now.
   The only customization one have to do to make a change is through
   register-use-preview which have now three options:

   - basic (default - same as Emacs-29- )
   - nil (simple preview with filtering, default and no confirm)
   - never (same as nil but without any preview)
   - t  (the improved version with navigation etc...)

I think the variable is being overloaded.  Shouldn't this just control
the usage of preview, not how interactive the behaviour of registers
are?

People seem to be confusing what it does already, since they think
that registers did not have preview before this change (and that being
the main feaature that they seem to be after -- not confirmation).
Which makes it hard to disucss this change, since it conflates
multiple topics into one thread.

"basic" seems to not be the same behaviour as in 29 (from the looks --
I cannot apply the patch), even Emacs 25 has register preview by
default, which is a useful feature, which seems to correspond to
'nil'.


The discussion of this change is strange and why people are upset, we
first change the behaviour silently and then try to justify the old
one.  It should be the other way round.  

Why is confirmation in registers so important to break old trusted
behaviour, when we already have a similar feature (bookmarks) that
those who wish to be "sure" can't use that instead?