Improved prompting and task type decisions for multi agent model

[joshhan619]

Planning_Agent.py:

• Improved the LLM prompts to get better answers from LLM
• Planning agent decides on best task type for TS Agent based on task description

TS_agent.py:

• Different behavior depending on task type
• If task type is classification, use a LLM to describe time series data
• If task type is forecasting, use time-series model (e.g. LTSM-bundle) to predict and then send description of predicted time-series to QA

QA_Agent.py:

• Changed prompt for LLM depending on task type

ltsm_inference.py:

• Changed the input of LTSM to use pretrained weights leading to improved forecasting results
• Added padding for the prompt data