note.txt

1. created annotation_mapping.compare_annotations to map between beat and downbeat times in score and performance
2. this is used by ops.anticipate2 which implements the interleave between controls and events using the interpolated mapping.
   this also omits events and controls that happen before the first beat or after the last beat, in which the interpolant is not defined.
3. tokenize.tokenize2 then converts a given list of files (MIDI score/performance and annotations) to interleaved sequences of length 1024,
   also making sure to pad the events (add rests so that the desired density is reached) 
4. finally, define tokenize-asap.py which passes all the ASAP data to tokenize2 and outputs into one file data/output.txt.

Potential problems:
- there are 31 annotation files between score and performance that don't match. 
- My solution was to choose the one with smaller length n and match together the first n many beats. But it's possible that the mismatch happens
in between, not at the end. If needed, I can manually remove these files later on.