Summary of important user-visible changes for datatypes 1.6.2:
--------------------------------------------------------------

 This patch release makes `ismissing` read its indicators as MATLAB does,
 and fixes `string` on big-endian machines and for text outside ASCII.

 ** `ismissing` on a `table` or `timetable` matches a `logical`, `single`
    or integer indicator only in variables of its own class and in `double`
    variables; it used to mark every numeric and logical variable, and a
    cell indicator such as `{int32(1), 2.4}` also marked 2.

 ** `missing` as an indicator marks the standard missing values of every
    class and of every table variable; it used to be refused or ignored.

 ** `NaN`, `NaT` and `<missing>` indicators now mark the missing elements of
    `duration`, `datetime` and `string` arrays.

 ** Text indicators on categorical data follow MATLAB: empty or blank text
    marks `<undefined>`, and in tables a `<missing>` or `""` indicator no
    longer does.  A blank indicator such as `' '` marks only cellstr
    elements holding that text, and a cell indicator no longer joins its
    texts into one.

 ** `calendarDuration.ismissing` takes an indicator, and tables take
    `calendarDuration` indicators, an Octave extension.

 ** Comparing strings with `<`, `<=`, `>` or `>=` no longer raises an index
    error when they hold characters outside ASCII.

 ** On big-endian machines, `strip` with no characters given removes
    whitespace again, strings with characters above U+00FF compare in the
    right order, and the test suite passes.


Backwards incompatible improvements:

 ** A `NaN` indicator in `ismissing` on a table no longer marks `NaN` in
    `duration` and `calendarDuration` variables, as in MATLAB; use
    `missing`, or a `NaN` of the variable's own class.
