Summary of important user-visible changes for datatypes 1.4.0:
--------------------------------------------------------------

 This release adds the `timetable` class and its event tables, and reworks the
 metadata every tabular class shares.  It breaks code written against the old
 `Properties` struct.

 ** New classes: `timetable`, `eventtable`, `eventfilter`, and the
    `datatypes.tabular` properties objects.

 ** `timetable` carries 37 methods: `retime` and `synchronize`, the joins, the
    grouping family, `rowfun`/`varfun`, the missing-data family, ordering and
    range predicates, and the variable-management family.

 ** Three file routes carry a timetable both ways: `timetable2csv`/
    `csv2timetable`, `timetable2ods`/`ods2timetable`, and the
    MATLAB-compatible `writetimetable`/`readtimetable`.  An event table
    survives only the OpenDocument route.

 ** `discretize`, `histcounts`, `groupcounts`, `groupfilter`, `groupsummary`
    and `grouptransform` are listed in INDEX.  They shipped in 1.3.4 but were
    absent from the catalogue, so `pkg describe` and `lookfor` did not find
    them.

 ** 27 improvements and 77 bug fixes, the largest groups in `retime`,
    `fillmissing`, `table`, `datetime` and `string`.


Backwards incompatible improvements:

 ** `Properties` is an object, not a struct, and no longer carries
    `VariableValues`.  Code calling `isstruct` on it, or reaching for
    `VariableValues`, must change.

 ** `csv2table` no longer takes the first column of a plain CSV as row names.
    A file read before this release gives a different table now.

 ** An unset `VariableUnits` or `VariableDescriptions` is empty rather than a
    cell of empty strings.

 ** `repelem` on a table or timetable requires both counts.  A count of zero
    is accepted and empties that dimension.

 ** `table2timetable` and `timetable2table` are methods of `table` and
    `timetable`.

 ** The three readers and the three writers take one set of name options.

 ** CSV and ODS files carry datetime and duration differently.

 ** `summary` prints its statistics block as the current MATLAB does, and
    reports the mean and the standard deviation.
