bug#50370: Fix bug in ispell-init-process error handling

[Eli Zaretskii <eliz@gnu.org>]

> Date: Fri, 3 Sep 2021 20:53:24 -0700
> From: Ian W <ian@wahbe.com>
> 
> This fixes the error handling in ispell-init-process. I encountered this bug, and also saw it mentioned here:
> https://mail.gnu.org/archive/html/auctex/2021-08/msg00007.html
> 
> The previous behavior involved the ispell process terminating
> prematurely (correct behavior, invalid input) and then calling
> ispell-accept-output. Because ispell-process had terminated,
> ispell-accept-output threw its own error, which prevented the underlying
> error from being displayed.
> 
> The bug resulted in the following interaction:
> ispell-word would print out:
> "Starting new Ispell process aspell with default dictionary...done"
> "ispell-accept-output: No Ispell process to read output from!"
> 
> The correct behavior is:
> "Starting new Ispell process aspell with default dictionary...done"
> "cond: Error: ~/.emacs.d/.local/etc/ispell/.pws: The language "nil" is not known. This is probably because: the
> file "/usr/local/Cellar/aspell/0.60.8/lib/aspell-0.60/nil.dat" can not be opened for reading.""

I'm not sure I agree that the "correct behavior" better than the
"incorrect".  The error message is not more self-explanatory, at
least.

Anyway, do you have a recipe for reproducing this problem?  The URL of
the AUCTeX discussion seems to say it happens every time?  I don't
think I understand what's going wrong to trigger the problem.

Thanks.