support sentencepiece tokenizer

tokenizer_config.json

@@ -20,6 +20,6 @@
     "special_tokens_map_file": null,
     "tokenizer_class": "BertJapaneseTokenizer",
     "word_tokenizer_type": "jumanpp",
-    "subword_tokenizer_type": "wordpiece",
+    "subword_tokenizer_type": "sentencepiece",
     "jumanpp_kwargs": {}
 }