First release since migration to the new Github organization htseq.

Binaries for Linux and OSX are provided on PyPI: https://pypi.org/project/HTSeq/#files.

As usual, installation with pip is recommended.

New features:

• Negative indices for StepVector (thanks to shouldsee for the original PR).
• htseq-count-barcodes counts features in barcoded SAM/BAM files, e.g. 10X Genomics
  single cell outputs. It supports cell barcodes, which result in different columns of
  the output count table, and unique molecular identifiers.
• htseq-count has new option -n for multicore parallel processing
• htseq-count has new option -d for separating output columns by arbitrary character
  (defalt TAB, , is also common)
• htseq-count has new option -c for output into a file instead of stdout
• htseq-count has new option --append-output for output into a file by appending to
  any existing test (e.g. a header with the feature attribute names and sample names)
• htseq-count has two new values for option --nonunique, namely fraction, which
  will count an N-multimapper as 1/N for each feature, and random, which will assign
  the alignment to a random one of its N-multimapped features. This feature was added by
  ewallace (thank you!).
• htseq-qa got refactored and now accepts an options --primary-only which ignores
  non-primary alignments in SAM/BAM files. This means that the final number of alignments
  scored is equal to the number of reads even when multimapped reads are present.

Testing improvements:

• Extensive testing and installation changes for Mac OSX 10.14 and later versions
• Testing Python 2.7, 3.6, 3.7, and 3.8 on OSX
• Testing and deployment now uses conda environments

Numerous bugfixes and doc improvements.

This is the last version of HTSEQ supporting Python 2.7, as it is unmaintained since Jan 1st, 2020. HTSeq will support Python 3.5+ from the next version.