mritc 0.6.0

• Package dependencies have been simplified. The lattice and misc3d
  packages have become optional (Suggests) dependencies rather than required
  ones, and tkrplot is not directly connected at all. Visualisation can now
  be performed by RNifti (an existing dependency through oro.nifti), which
  has more easily met system requirements than misc3d.
• A suite of tests has been added to the package. tinytest and covr are
  now suggested dependencies for testing and code-coverage support.
• plot.mritc() gains a method argument for choosing the visualisation
  backend: the new default, "RNifti", uses RNifti::view() and requires
  only packages already needed elsewhere by mritc; the previous behaviour is
  still available via method="misc3d", which uses the Tk-based
  misc3d::slices3d() viewer and requires the misc3d and tkrplot packages
  to be installed.
• Added a verbose argument to mritc(), passed through to the underlying
  fitting methods, to allow progress messages to be suppressed.
• Fixed a bug where mritc(..., method="MCMCsubbias") used the wrong
  spatial neighbourhood construction (sub=FALSE instead of sub=TRUE),
  affecting results for that method.
• Fixed a bug in mritc.pvhmrfem() where argument checks on mu, sigma
  and err were not actually performed, which could lead to a downstream
  segfault for invalid inputs.
• Fixed a bug in writeMRI() involving incorrect vectorised use of
  ifelse(), which could fail for some input object types.
• Class labels are now assigned deterministically when tied
  (print.mritc(), summary.mritc(), plot.mritc() and measureMRI() all
  used max.col() without breaking ties consistently, which could change
  reported results from run to run).
• The package citation style has been updated to match current requirements.