對tokenizer增加special tokens使其能被.apply_chat_template正確轉換

作者您好，之前為chatglm3[添加chat_template](https://huggingface.co/THUDM/chatglm3-6b/discussions/22)後發現，使用`tokenizer.apply_chat_template`轉換成`input_ids`時會因為`tokenizer`不認識`special token`而導致錯誤的轉換結果，因此提出此PR進行修正。

詳細比較與修改請參閱gist:
https://gist.github.com/p208p2002/adf1d4235b9567227d01315beb4b210e

此PR包含 :
1. 對`tokenizer`增加四個special token ，使他們能夠被正確轉換。
```python
tokenizer.add_tokens("<|user|>")
tokenizer.add_tokens("<|assistant|>")
tokenizer.add_tokens("[gMASK]")
tokenizer.add_tokens("sop")
```
2. 移除 `chat_template` 中額外空白

貢獻:
此PR能夠使 chatglm3 的 tokenizer 正確支援 `tokenizer.apply_chat_template`方法。

special_tokens_map.json

@@ -0,0 +1 @@
+{}

tokenizer_config.json

@@ -1,13 +1,52 @@
 {
-  "name_or_path": "THUDM/chatglm3-6b",
-  "remove_space": false,
-  "do_lower_case": false,
-  "tokenizer_class": "ChatGLMTokenizer",
-  "chat_template": "{% for message in messages %}{% if loop.first %}[gMASK]sop<|{{ message['role'] }}|> \n {{ message['content'] }}{% else %}<|{{ message['role'] }}|> \n {{ message['content'] }}{% endif %}{% endfor %}{% if add_generation_prompt %}<|assistant|>{% endif %}",
+  "added_tokens_decoder": {
+    "64790": {
+      "content": "[gMASK]",
+      "lstrip": false,
+      "normalized": true,
+      "rstrip": false,
+      "single_word": false,
+      "special": false
+    },
+    "64792": {
+      "content": "sop",
+      "lstrip": false,
+      "normalized": true,
+      "rstrip": false,
+      "single_word": false,
+      "special": false
+    },
+    "64795": {
+      "content": "<|user|>",
+      "lstrip": false,
+      "normalized": true,
+      "rstrip": false,
+      "single_word": false,
+      "special": false
+    },
+    "64796": {
+      "content": "<|assistant|>",
+      "lstrip": false,
+      "normalized": true,
+      "rstrip": false,
+      "single_word": false,
+      "special": false
+    }
+  },
   "auto_map": {
     "AutoTokenizer": [
-      "tokenization_chatglm.ChatGLMTokenizer",
+      "THUDM/chatglm3-6b--tokenization_chatglm.ChatGLMTokenizer",
       null
-      ]
-  }
+    ]
+  },
+  "chat_template": "{% for message in messages %}{% if loop.first %}[gMASK]sop<|{{ message['role'] }}|>\n {{ message['content'] }}{% else %}<|{{ message['role'] }}|>\n {{ message['content'] }}{% endif %}{% endfor %}{% if add_generation_prompt %}<|assistant|>{% endif %}",
+  "clean_up_tokenization_spaces": false,
+  "do_lower_case": false,
+  "eos_token": "</s>",
+  "model_max_length": 1000000000000000019884624838656,
+  "pad_token": "<unk>",
+  "padding_side": "left",
+  "remove_space": false,
+  "tokenizer_class": "ChatGLMTokenizer",
+  "unk_token": "<unk>"
 }