Add Baseline Agents and Benchmark Configurations

[AnkitaNaik]

Title
Integrate baseline agents and benchmarking configurations

Description
Summary: Expand benchmarking beyond CUGA to support baseline comparisons.

Scope

Support

• Agents
  -- CUGA
  -- LangGraph ReAct
  -- AppWorld CodeAct

• Models
  -- GPT-OSS-120B
  -- GPT-OSS-20B
  -- Mistral Medium 3.5
  -- Gemma 4 31B
  -- Nematron 3

Acceptance Criteria

• Multiple agents supported
• Multiple models configurable
• Baseline comparisons generated