Set defaults for optional parameters in ChatOCIModelDeploymentVLLM

[qiuosier]

The ChatOCIModelDeploymentVLLM class contains inference parameters for vLLM APIs. However, updates on vLLM may remove some parameters, causing breaking change even if users do not use those parameters.

For example, the early_stopping parameter exist for v0.6.2 but it is removed since v0.6.3.

In the ChatOCIModelDeploymentVLLM class there is already logic to include parameters in the API request only if they are not None. However, since some of the parameters has a default value instead of None, they are added to the API request even though users did not intend to use them.

This PR sets the defaults of the optional parameters to None, so that they will not be added to the API request unless users specified them explicitly.

[github-actions]

📌 Cov diff with main:

📌 Overall coverage:

[VipulMascarenhas]

@qiuosier I thought "autogen-agentchat~=0.2" would restrict it to 0.2.x, but seems it install 0.4.0 version for python 3.10+ but not for 3.9. The autogen and tf2onnx tests should be fixed if we change it to "autogen-agentchat<0.4".

[qiuosier]

Thanks @VipulMascarenhas . I created a new PR for the autogen issue. @mrDzurb please suggest whether to merge this PR since your are testing the tool calling vllm image.

[github-actions]

📌 Cov diff with main:

📌 Overall coverage: