5' Sequence and fragment GC-content bias correction

[timbitz]

This is another very large PR which implements bias correction methods:

• 5' Sequence bias correction method by Hansen et al. NAR 2010 for both TPM and PSI calculations
• Fragment GC-content bias for PSI calculations: Background GC-content built into Whippet index for each transcript, expression weighted after isoform-level EM and then edge counts are corrected.
• DefaultCounter/DefaultBiasMod which are simply type-stable wrappers for Float64

[codecov[bot]]

Codecov Report

Merging #59 into 0.6-genelength-long will increase coverage by 0.49%. The diff coverage is 46.69%.

@@                   Coverage Diff                   @@
##           0.6-genelength-long      #59      +/-   ##
=======================================================
+ Coverage                51.62%   52.11%   +0.49%     
=======================================================
  Files                       17       18       +1     
  Lines                     2127     2318     +191     
=======================================================
+ Hits                      1098     1208     +110     
- Misses                    1029     1110      +81

Impacted Files	Coverage Δ
src/ebi.jl	80% <ø> (+80%)	:arrow_up:
src/graph.jl	93.51% <100%> (+1.76%)	:arrow_up:
src/sgkmer.jl	80% <100%> (+1.42%)	:arrow_up:
src/io.jl	33.14% <11.11%> (-0.1%)	:arrow_down:
src/quant.jl	44.33% <31.5%> (-2.91%)	:arrow_down:
src/events.jl	56.94% <45.83%> (ø)	:arrow_up:
src/bias.jl	49.09% <49.09%> (ø)
src/reads.jl	30.18% <55%> (+16.67%)	:arrow_up:
... and 1 more

---

Continue to review full report at Codecov.

Legend - Click here to learn more Δ = absolute <relative> (impact), ø = not affected, ? = missing data Powered by Codecov. Last update a5485e2...3a5b94e. Read the comment docs.