bug#45617: <query-replace> loses the edit region. Works in 23.3, broke in 26.3

[Juri Linkov]

> For my two cents, what you write sounds like the real problem is the
> minibuffer failing to restore on exit. A fix to that would be preferable to
> the patch. After all, who knows what else may affected by the minibuffer!
> I'd be happy to test this alternative.

The question we need to answer: should this fix affect all other uses
of the minibuffer?  So for any command that operates on the active region
and asks its arguments from the minibuffer, when you copy the text
from the original buffer to the minibuffer and thus change its region,
should quitting the minibuffer restore the original region for
all such commands?  This means restoring the original mark and point.