Remove dead snapshot.mode override in DebeziumIO.getRecordSchema

getRecordSchema copied the connector configuration, set snapshot.mode on
the copy, then passed the original unmodified map to getOneRecord. The
override therefore never took effect, and schema inference has always run
with the connector's configured snapshot mode.

Drop the dead copy rather than start honouring it. getOneRecord derives
the schema by sampling an actual data record, so a schema-only snapshot
would emit nothing and schema inference would fail with "could not fetch
database schema". No behaviour change.

Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>

Author: tkaymak

sdks/java/io/debezium/src/main/java/org/apache/beam/io/debezium/DebeziumIO.java

@@ -36,7 +36,6 @@
 import org.apache.beam.sdk.values.PCollection;
 import org.apache.beam.vendor.guava.v32_1_2_jre.com.google.common.base.Joiner;
 import org.apache.beam.vendor.guava.v32_1_2_jre.com.google.common.collect.Lists;
-import org.apache.beam.vendor.guava.v32_1_2_jre.com.google.common.collect.Maps;
 import org.apache.kafka.connect.source.SourceConnector;
 import org.apache.kafka.connect.source.SourceRecord;
 import org.checkerframework.checker.nullness.qual.Nullable;
@@ -313,9 +312,9 @@ protected Schema getRecordSchema() {
           new KafkaSourceConsumerFn.OffsetTracker(
               new KafkaSourceConsumerFn.OffsetHolder(null, null, 0)));
 
-      Map<String, String> connectorConfig =
-          Maps.newHashMap(getConnectorConfiguration().getConfigurationMap());
-      connectorConfig.put("snapshot.mode", "schema_only");
+      // Deliberately runs with the connector's configured snapshot mode: schema inference samples
+      // an actual data record, which a schema-only snapshot ("no_data", formerly "schema_only")
+      // would never emit.
       SourceRecord sampledRecord =
           fn.getOneRecord(getConnectorConfiguration().getConfigurationMap());
       fn.reset();