bug#52053: 29.0.50; Nonsensical button "C-x C-f" in scratch buffer

[Gregory Heytings]

> Yes, but that's somewhat blunt a weapon, I think.  My idea was to maybe 
> let the caller decide whether this is wanted or not, if we are going to 
> use substitute-command-keys in way more contexts than it was originally 
> meant to.

AFAICS, the options are:

1. Detect inside substitute-command-keys whether the string is for a 
*Help* buffer.  That's what the patch does, it's the simplest solution 
from the point of view of those who call s-c-k.

2. Add an optional parameter to s-c-k to indicate whether the string is 
for a non-*Help* buffer.  That's a reasonable alternative, but there are 
~450 calls to s-c-k in the trunk, each would have to be checked.

3. Add a variable and let-bind it around the calls to s-c-k to indicate 
when the string is for a non-*Help* buffer.

4. Let-bind help-link-key-to-documentation to nil around the calls to 
s-c-k to indicate that the calls to help-mode--add-function-link should be 
skipped.  ISTM that this is not generic enough, if in six months we had 
another similar feature it would be necessary to let-bind two variables 
around calls for non-*Help* buffers, and so forth.