New features:
- core.shortname switch which results in short file names for figures
- field.colorbar switch to remove colorbar (might be useful when several
  plots share the same colorbar)
- info subcommand understands +dimensional switch
- size of rprof figures is now adapted to the number of plots
- the field subcommand now uses the same ',.-'-separated list mechanism
  to specify how fields should be spread across subplots and figures
- remove shrinckb options as the size of colorbars is now automatic
- the field subcommand learned the timelabel switch
- automatic list of configuration options in online doc
- Python 3.8 support

Bug fixes:
- _Fields.__delitem__ is robust against asking to delete a non existing
  field
- views over _Snaps no longer return EmptyStep
- slices of 3D vector fields are fixed
- do not load mplstyle when isolated
- fallback to local version of mplstyle if not found
- consistent capitalization of config help

Performances:
- _Fields.__contains__ no longer relies on __iter__, it directly checks
  for the existence of the field data

API changes:
- _Steps and _Snaps implement __delitem__, useful to release memory
- new geom.at_z method returns iz closest to a given z
- StagyyData.sfields are now labeled
- InvalidTimestepError is a KeyError
- InvalidSnapshotError, also a KeyError, is raised when accessing
  a non-existing snapshot instead of an EmptyStep; the latter no
  longer exist
- _Snaps._bind is now marked as an internal method
- StagyyData attributes collected_fields, files, and binfiles_set no
  longer in public API
- StagyyData constructor path argment is now optional and defaults to
  core.path
- StagyyData.nfields_max is a property, and should be set to None to
  lift the limit on the number of fields in memory
- more values are considered as truthy for envrionment variable, see
  documentation
- remove misc.fmttime
- remove public step.isnap setter
- remove (_Steps|_Snaps).last

Internals:
- docstring format is now checked by flake8
- https URLs in doc
- cleanup of travis configuration
- new makefile recipe to access venv in Jupyter
- most recent versions of pip support editable install again and can be
  used to build the venv for the development version