Fix ollama client TODOs

[peterychang]

What feature would you like to be added?

Several TODOs in ollama client:

• Add model families to directly supported models
• Deduplicate code between openai, azure, and ollama model clients
• Add argument checking
• Handle message source parameters
• Ensure token counting logic is correct
• Ensure constants (eg: done_reason) are still correct
• Check whether multiple content, images, and/or tools can/should exist within a single message
• Confirm streaming mode works properly
• Implement logprobs logic when it's added to the ollama client (see: Provide logits or logprobs in the API ollama/ollama#2415)

Why is this needed?

necessary to fully support the client

[codelion]

You can try using optillm - https://github.com/codelion/optillm it already supports logits for local inference and implements several test-time-compute techniques that can benefit autogen.

[victordibia]

@peterychang , are these fixed in #5595 ?

[peterychang]

@peterychang , are these fixed in #5595 ?

No, these are most of the things that weren't properly implemented or tested in the initial client release