new for version 1.35.1
======================
* gsea(), vulcan.annotate() and vulcan.pathways() now resolve their 'method'
  argument with match.arg(). Called without naming a method, as their own
  documentation describes, each one reached an if() on a vector of length
  greater than one, which is an error in R >= 4.2
* null_gsea() no longer fails when a permutation places the whole set on
  elements of weight zero, which left the running score undefined
* gsea() accepts a null distribution given as a named list keyed by set
  size. The test that was meant to detect one named a class that nothing
  ever produced, and could itself error when handed a matrix
* corr2p() asks for the upper tail directly instead of subtracting from
  one, and so no longer returns exactly 0 for strong correlations
* kmgformat() returns "0" rather than "00" for zero, which pretty() puts on
  most axes
* val2col() centres the data before measuring the extent of the colour
  scale, no longer fails on data with a small range, and no longer returns
  the smallest value without a colour
* average_fragment_length() corrects an off-by-one: correlateReads() counts
  delays from 0, so the reported length was one base too long
* assay() is imported from SummarizedExperiment rather than relied upon
  being attached by something else
* DESCRIPTION uses Authors@R

new for version 1.9.3
=====================
* Minor udpates in the package description and vignette