Add RNA squiggle model

Bermuda

• Temperature corrected R10 model (rgrgr_r10)
• New R9.4 squiggle model with improved dwell correction
• Allow read translocation rate to vary in signal-to-squiggle mapping

release-1.4.0

• Direct correction for homopolymers
• Temperature adjustment for improved accuracy (rgrgr_r94, rgrgr_r941 and rgrgr_r10 models only).
• Module to detect event boundaries without calling.
• Remove some old models

Regress code from C11 to C99 to allow compilation on Mac OS and older Linux platforms that don't support aligned_alloc in their C library.

Python interface to support Read Until applications.

Initial support for Recurrent Neural network -- Random Field models

A recurrent neural network parameterizes a conditional random field over the samples for labels (A, C, G, T or "blank") and is then decoded to produce a base call.

Initial support for sequence-to-squiggle prediction

• Improved rgrgr models
• Ability to write output to named file
• Rename argument '--outformat' to '--format'

• release-1.1.0
  • Implementation of raw data 'pirate' networks (rGr and rgrgr).
  • Stand-alone event detection.
  • Stay penalty to allow ratio of insertions to deletions to be changed.
  • Local (Smith-Waterman like) penalty on edges of basecall to remove noisy bases.
  • Improvements to error reporting.

• Change licence to MPL.
• Increased number for assertions.
• A few stability and bug fixes.