Include reasoning tokens and token streaming for anthropic LLMs

[damithsenanayake]

Missing Thoughts and Unable to Token Stream with Claude Models
To leverage models put forward by Anthropic through model garden, we utilise the anthropic_llm.py where Claude models are defined. However, this does not allow for token streaming or forwarding reasoning tokens to the UI.

Solution
Implement:

• parsing Google GenAI's Thinking Config into Anthropic's Thinking Config to be passed along to anthropic API
• processing the anthropic API messages into response parts compatible with ADK
• allow for custom configurations (e.g., beta headers) in Anthropic API

Alternatives

• LiteLLM to process thinking tokens (Thought / Reasoning messages missing in events #2982)

Proposed Fix
PR created #3070

[damithsenanayake]

@boyangsvl thanks for looking into this. happy to make any changes if required before merging.