Simplify redisplay-x/separate_textual_runs

[Aidan Kehoe]

 Ar an t-aonú lá is triochad de mí Márta, scríobh Stephen J. Turnbull: 

 > Aidan Kehoe writes:
 > 
 >  > Thanks! Please profile the two (with etc/HELLO at least working for
 >  > the CJK languages) and post comparisons, separate_textual_runs is
 >  > called quite a lot.
 > 
 > With all due respect, I don't think this is very useful at this point.
 > 
 > First, in light usage I perceive no slowdown, and don't expect a
 > significant speedup, either.

Perception’s not relevant if we have profiling results. 

 > Second, before optimizing separate_textual_runs, which has always been
 > a one-pass O(text length) algorithm, and thus hard to beat (or screw
 > up), redisplay (and therefore separate_textual_runs, I suspect) seems
 > to be called multiple times for an awful lot of changes; saving whole
 > calls the separate_textual_runs would be more helpful.

And the chances of someone getting to that now, as opposed to when we get
some more tuits, are immeasurably small, while the chances of Olivier
getting some useful results out of profiling code which he currently
understands, are more realistic.

My last changes to separate_textual_runs eliminated unnecessary assignments
during the loop that had been there since the Mule merge. We don’t have a
good record on this, and doing something to improve performance is
preferable to the tuits approach. 

 > There's also an issue that Xft fonts don't seem to be properly cached
 > (although on the server side it may not be implemented, which is a
 > fact for PROBLEMS, and the chatter I'm seeing from Xft about
 > initializing fonts may be due to me putting the debug printfs in the
 > wrong place).

-- 
On the quay of the little Black Sea port, where the rescued pair came once
more into contact with civilization, Dobrinton was bitten by a dog which was
assumed to be mad, though it may only have been indiscriminating. (Saki)