[Feat] Provider-Specific Param Support in Langchain

[reyna-abhyankar]

Different provider classes (e.g., ChatOllama) may have their own format for invocation params (i.e. keyword arguments that we pass to the endpoint) that we would have to parse. Currently, we treat collect this using _get_invocation_params() which may not be exhaustive or correct in some cases (see #10 for ChatOllama). However, _get_ls_params() leaves out potential arguments.

May need to write provider-specific code to collect the parameters that can be passed directly to the LiteLLM endpoint.