bug#16800: 24.3; flyspell works slow on very short words at the end of big file

[Eli Zaretskii <eliz@gnu.org>]

> From: Aleksey Cherepanov <aleksey.4erepanov@gmail.com>
> Date: Wed, 19 Feb 2014 00:56:45 +0400
> 
> I faced a problem editing my big .org file (2mb+) with flyspell-mode
> enabled. I edit it every day, regularly mistype and get words of one
> or two letters that are wrong in Russian and cause flyspell work slow.
> 
> This one-liner produces "good" file to reproduce the bug.
> perl -e 'print(((join " ", ("met and") x 10) . "\n") x 30000)' > t.txt
> 
> Typing "nd" at the end of file gives a huge pause even on a fast
> computer. But "mw" or "md" does not give pauses because they are not
> substrings in this file. It is repeatable with emacs -Q.

This seems to be due to the Flyspell's feature of recognizing
duplicates of mis-spelled words, and, if found, highlighting such
duplicates in a different face.  If you customize the variable
flyspell-duplicate-distance to some small value (or even zero), the
delay goes away.  Evidently, with the default value of -1, Flyspell
searches all the way to the beginning of the giant buffer, looking for
a duplicate of "nd".

Interestingly, I don't see this when the speller is Ispell, but I do
see it with Hunspell.  Not sure how using Ispell avoids this problem.