[BUG] LiteLLM provider does not handle reasoning fields #447

Describe the bug
LiteLLMHandler does not process reasoning_content or reasoning fields from the streaming delta. When using
reasoning models (e.g. DeepSeek-R1, QwQ) routed through a LiteLLM proxy, the thinking/reasoning output is
silently dropped and never displayed to the user.

To Reproduce
Steps to reproduce the behavior:

1. Set up a LiteLLM proxy server pointing to a reasoning model (e.g. DeepSeek-R1 or QwQ-32B)
2. Configure Zoo Code to use the LiteLLM provider with that model
3. Send any message and observe the response

Expected behavior
The reasoning/thinking content should be displayed in the UI before the final response, as it is when using
DeepSeek or other reasoning-capable providers directly.

Screenshots
N/A

Video
N/A

What version of zoo are you running
3.56.0

Additional context
BaseOpenAiCompatibleProvider (used by DeepSeek, Requesty, Unbound, etc.) already handles this correctly by
checking delta.reasoning_content and delta.reasoning in the stream loop. However, LiteLLMHandler overrides
createMessage independently and does not include this logic, causing the field to be silently ignored.