chaining- rerun with new version

[jackhump]

be strict,

chain_samples.py --dun-merge-5-shorter --max_fuzzy_junction=0 --max_3_diff=30

[jackhump]

scratch that.

use dev version of cupcake: https://github.com/Magdoll/cDNA_Cupcake/tree/dev

first cupcake collapse (without dun-merge) to collapse 5' degraded transcripts and count abundances

Then do chaining, again without dun-merge

Then do a final round of collapsing (without dun-merge) to now collapse 5' degraded (but full length in that sample) reads into hopefully the longest read across all transcripts.