fix: Removes redundant token calculations and updates dependencies

Eliminates unnecessary pre-calculation of token limits and recalculation of max tokens
across multiple app runners, simplifying the logic for prompt handling.

Updates tiktoken library from version 0.8.0 to 0.9.0 for improved tokenization performance.

Increases default token limit in TokenBufferMemory to accommodate larger prompt messages.

These changes streamline the token management process and leverage the latest
improvements in the tiktoken library.

Fixes potential token overflow issues and prepares the system for handling larger
inputs more efficiently.

Relates to internal optimization tasks.

Signed-off-by: -LAN- <laipz8200@outlook.com>

api/core/agent/cot_agent_runner.py

@@ -104,7 +104,6 @@ class CotAgentRunner(BaseAgentRunner, ABC):
 
             # recalc llm max tokens
             prompt_messages = self._organize_prompt_messages()
-            self.recalc_llm_max_tokens(self.model_config, prompt_messages)
             # invoke model
             chunks = model_instance.invoke_llm(
                 prompt_messages=prompt_messages,

api/core/agent/fc_agent_runner.py

@@ -84,7 +84,6 @@ class FunctionCallAgentRunner(BaseAgentRunner):
 
             # recalc llm max tokens
             prompt_messages = self._organize_prompt_messages()
-            self.recalc_llm_max_tokens(self.model_config, prompt_messages)
             # invoke model
             chunks: Union[Generator[LLMResultChunk, None, None], LLMResult] = model_instance.invoke_llm(
                 prompt_messages=prompt_messages,