Change GPU efficient textcat to use CNN, not BOW

If you generate a config with a textcat component using GPU
(transformers), the defaut option (efficiency) uses a BOW architecture,
which does not use tok2vec features. While that can make sense as part
of a larger pipeline, in the case of just a transformer and a textcat,
that means the transformer is doing a lot of work for no purpose.

This changes it so that the CNN architecture is used instead. It could
also be changed to be the same as the accuracy config, which uses the
ensemble architecture.

spacy/cli/templates/quickstart_training.jinja

@@ -221,10 +221,16 @@ no_output_layer = false
 
 {% else -%}
 [components.textcat.model]
-@architectures = "spacy.TextCatBOW.v2"
+@architectures = "spacy.TextCatCNN.v2"
 exclusive_classes = true
-ngram_size = 1
+nO = null
-no_output_layer = false
+
+[components.textcat.model.tok2vec]
+@architectures = "spacy-transformers.TransformerListener.v1"
+grad_factor = 1.0
+
+[components.textcat.model.tok2vec.pooling]
+@layers = "reduce_mean.v1"
 {%- endif %}
 {%- endif %}
 
@@ -252,10 +258,16 @@ no_output_layer = false
 
 {% else -%}
 [components.textcat_multilabel.model]
-@architectures = "spacy.TextCatBOW.v2"
+@architectures = "spacy.TextCatCNN.v2"
 exclusive_classes = false
-ngram_size = 1
+nO = null
-no_output_layer = false
+
+[components.textcat_multilabel.model.tok2vec]
+@architectures = "spacy-transformers.TransformerListener.v1"
+grad_factor = 1.0
+
+[components.textcat_multilabel.model.tok2vec.pooling]
+@layers = "reduce_mean.v1"
 {%- endif %}
 {%- endif %}