feat(model): Support model cache and new Agentic Workflow Expression Language(AWEL)

[fangyinc]

Model cache

• Cache model output in memory.
• Cache model output in local rocksdb.

Agentic Workflow Expression Language(AWEL)

• Common workflow Operators
• Implements LocalWorkflowRunner
• Integrate model serving and model result cache with AWEL in base_chat.py

Note:

DB-GPT uses local disk cache by default. You need to install related dependencies by command:

pip install -e ".[cache]"

If there are no dependencies of disk cache, DB-GPT will uses memory cache.

Closes #788

[Aries-ckt]

neat code!