Add "pooler-output" to sinequa.metadata.json

Following the Pooling strategy changes for Sinequa model packagings, we enforce for this vectorizer the use of the model 'pooler_output' instead of the strategy defined in '1_Pooling/config.json'

sinequa.metadata.json

@@ -1,5 +1,6 @@
 {
 	"score-scaling-factor": 0.1,
 	"tanh-activate": true,
-	"mask-padding": true
+	"mask-padding": true,
+    "pooler-output": true
 }