Fix TypeError when adjusting with preloaded training dataset (#1679)

See issue #1678 for an explanation of the problem

### What kind of change does this PR introduce?

Returns an empty list if a reduced dimension is not in the dataset
chunks in xclim/sdba/base.py map_blocks when checking for badchunks

### Does this PR introduce a breaking change?
No.

AUTHORS.rst

@@ -41,3 +41,4 @@ Contributors
 * Maliko Tanguy <[email protected]> `@malngu <https://github.com/malngu>`_
 * Christopher Whelan `@qwhelan <https://github.com/qwhelan>`_
 * Dante Castro <[email protected]> `@profesorpaiche <https://github.com/profesorpaiche>`_
+* Sascha Hofmann <[email protected]> `@saschahofmann <https://github.com/saschahofmann>`_

CHANGES.rst

@@ -13,6 +13,7 @@ Announcements
 Bug fixes
 ^^^^^^^^^
 * Fixed an bug in sdba's ``map_groups`` that prevented passing DataArrays with cftime coordinates if the ``sdba_encode_cf`` option was True. (:issue:`1673`, :pull:`1674`).
+* Fixed bug (:issue:`1678`, :pull:`1679`) in sdba where a loaded training dataset could not be used for adjustment
 
 Internal changes
 ^^^^^^^^^^^^^^^^

xclim/sdba/base.py

@@ -611,7 +611,7 @@ def _map_blocks(ds, **kwargs):  # noqa: C901
                     {
                         dim: chunks.get(dim)
                         for dim in reduced_dims
-                        if len(chunks.get(dim)) > 1
+                        if len(chunks.get(dim, [])) > 1
                     }
                 )
                 if badchunks: