Update TODO.

Signed-off-by: Stefan Westerfeld <stefan@space.twc.de>

src/TODO

 speed detection possible improvements:
 - refine with 128 sync step instead of 256?
-- use bigger initial region (50s) --detect-speed-harder
-- use window hann instead of hamming for sync bits (or all bits)?
-- remove lrint and QInterpolator?
+- grid size adjustment
+- remove lrint?
 - detect silence instead of total volume?
 - split big region into two smaller ones?
 - connected search on even smaller regions?
 - reorder memory access in compare to be more cache efficient
-- n-best search?
-- soft sync?
 
 possible improvements:
 - dynamic bit strength