bug#56682: Fix the long lines font locking related slowdowns

[Gregory Heytings]

> > When I visit addons.json (from my Firefox directory), which is a single 
> > line of (only) 19540 characters, and type `C-n' and hold down both 
> > keys, the cursor stops moving almost immediately (on the third visual 
> > line) and after letting go of the keys and waiting several seconds, 
> > point jumps down the buffer to its actual position; likewise with 
> > `C-p'. This is on master with -Q, and is no different from Emacs 28 
> > (the only difference I notice is that in master the buffer is wrongly 
> > fontified from position 8235 to the end, while in 28 the entire buffer 
> > is correctly fontified).  Is this expected?  (When I enable 
> > so-long-mode in the buffer, holding down `C-n' or `C-p' produces no 
> > delay, both in master and 28.)
>
> That's not expected, no, and rather surprising because it would 
> contradict what we've seen so far.  Can you post that file somewhere? 
> The only possible reason I can see ATM is that the file contains Arabic 
> characters, which are (and will remain) slow.

I forgot to mention: did you try after disabling show-paren-mode?  It is 
known to slow down redisplay at the beginning and end of the buffer, and 
will eventually be automatically turned off in these buffers.