use a different megahal implementation, which seems to be better at saving the brain.

note that nothing megahal is shipped with the bot now, including
train files. this bot will expect a ./megahal.trn, and will write
to ./megahal.brn (in addition to a slew of other ./megahal.*
files).

implementation is from http://megahal.alioth.debian.org/

.gitignore

@@ -8,3 +8,4 @@ dr.botzo.cfg
 nbproject
 parser.out
 parsetab.py
+megahal.*