ensure that -i3 pads to full block size

otherwise we lose the nominal burst error correction capability. if a feature to remove the padding existed, it'd probably be based on shortenings to RS codes (i.e. we know exactly which values must be zero judging just from the block length, so they don't count towards the errors, allowing us to get better parity to data ratio)

README.md

@@ -37,9 +37,6 @@ A rough outline of some development-related topics below.
 
 ## Roadmap
 
-High priority:
-- Should probably not pad to the full interlacing block size.
-
 Low priority:
 - Speed up trans2D/trans3D in jmode.c (might need native code).