Remove DictEncoder dtor checking in Parquet writer (#10445)

Summary:
Below exception is noticed when running spark_aggregation_fuzzer_test with
#9559, which writes Velox vectors into Parquet files for Spark to read.

```
velox/dwio/parquet/writer/arrow/Encoding.cpp:513:  Check failed: buffered_indices_.empty()
./velox/functions/sparksql/fuzzer/spark_aggregation_fuzzer_test
```
This PR follows apache/arrow@02ad5ae to remove this check.

Pull Request resolved: #10445

Reviewed By: kagamiori

Differential Revision: D59814851

Pulled By: mbasmanova

fbshipit-source-id: a72e6f7608d8bb1763e11e57ef1e5b8daa40f9cf

velox/dwio/parquet/writer/arrow/Encoding.cpp

@@ -505,9 +505,7 @@ class DictEncoderImpl : public EncoderImpl, virtual public DictEncoder<DType> {
         dict_encoded_size_(0),
         memo_table_(pool, kInitialHashTableSize) {}
 
-  ~DictEncoderImpl() override {
-    DCHECK(buffered_indices_.empty());
-  }
+  ~DictEncoderImpl() = default;
 
   int dict_encoded_size() const override {
     return dict_encoded_size_;