BUG: Fix default for post_tfm_kwargs in config classes

Default should be None, not an empty dict.
Setting it to an empty dict causes logic in core/eval
to instantiate a PostProcess transform
with majority_vote = False and min_segment_dur = None,
so that we end up applying post-processing that does nothing,
and the computed metric is the same as
without "post-processing".
This sets the default to None,
which fixes the issue.

src/vak/config/eval.py

@@ -127,7 +127,7 @@ class EvalConfig:
     post_tfm_kwargs = attr.ib(
         validator=validators.optional(are_valid_post_tfm_kwargs),
         converter=converters.optional(convert_post_tfm_kwargs),
-        default={},  # empty dict so we can pass into transform with **kwargs expansion
+        default=None,
     )
 
     # optional, data loader

src/vak/config/learncurve.py

@@ -74,5 +74,5 @@ class LearncurveConfig(TrainConfig):
     post_tfm_kwargs = attr.ib(
         validator=validators.optional(are_valid_post_tfm_kwargs),
         converter=converters.optional(convert_post_tfm_kwargs),
-        default={},  # empty dict so we can pass into transform with **kwargs expansion
+        default=None,
     )