Document derived-bounds support, including whatsnew. (#6555)

Author: pp-mo

docs/src/whatsnew/latest.rst

@@ -49,6 +49,12 @@ This document explains the changes made to Iris for this release
 #. `@pp-mo`_ made it possible for the reference surfaces of derived coordinates, like orography, to be lazy.
    (:pull: 6517).
 
+#. `@HGWright`_ and `@pp-mo`_ enabled correct loading and saving of the bounds of CF
+   parametric coordinates (that is, Iris derived coordinates).  This was previously
+   incorrect.  However the fix is opt-in, controlled by the ``derived_bounds`` flag in
+   the :data:`iris.FUTURE` object, to avoid breaking existing code.
+   (:issue:`3678`, :pull:`6481`, :pull:`6540`)
+
 
 🐛 Bugs Fixed
 =============

lib/iris/__init__.py

@@ -217,7 +217,12 @@ def __init__(
             may need to defend against floating point precision issues where
             you didn't need to before.
         derived_bounds : bool, default=False
-            When True, uses the new form for deriving bounds with the load.
+            When ``True``, uses the correct CF rules for bounds of derived coordinates
+            for both loading and saving NetCDF.  This requires that these must be linked
+            via a separate "formula_terms" attribute on the bounds variable.
+            If ``False``, bounds are only linked with a "bounds" attribute, though this
+            is strictly incorrect for CF >= v1.7.
+            See `here in CF <https://cfconventions.org/Data/cf-conventions/cf-conventions-1.7/cf-conventions.html#cell-boundaries>`_.
         lam_pole_offset : bool, default=False
             When True, saving a cube on a "Limited Area Model" (LAM) domain
             to a PP file will set the pole longitude (PP field ``bplon``) to