[TxMt] word completion lookup

[Caio Chassot]

> I have changed it now so that it does does mix in the word after the 
> caret, with those in front (loosely based on proximity) -- watch my 
> appcast :)

I've been using this new version for a few days now and it's highly 
unpredictable. Can you reveal what algorithm you're using?

What I had in mind was prioritizing matches like this:

-0
+0
-1
+1
-2
+2

Where - indicates before the caret, and + after, and the number is the 
number of lines offset from the current line. Within those groups you 
just order by position in the line.