bug#81537: 32.0.50; Fixing icomplete-in-buffer requiring users to advice-add

>> The simplest way to do this would be to use the new
>> 'display-buffer-default-alist':
>
> Thanks, though, this is inefficient -- I think we should avoid calling
> all the code in minibuffer-completion-help if we don't want to display
> *Completions*.

I still don't understand why should we suppress displaying of
*Completions* after the user types 'C-M-i' (completion-at-point)
whose primary purpose is to display the *Completions* window
like 'TAB' (minibuffer-complete) does in the minibuffer
even when icomplete-mode is enabled.

Can we step back and rethink this feature?  When doing it right,
displaying inline icomplete completions should not require typing
a special key.  It should show completions immediately like
icomplete-mode already does in the minibuffer.

This will require implementing global-icomplete-in-buffer-mode
and per-buffer icomplete-in-buffer-mode with post-command-hook.
I could try to do this.