Ollama fix handle incomplete JSON chunks in stream

[matmorel]

Addresses one of the issues raised in #686

Problem

Ollama LLM can return chunk in parts when it's too long.

Solution

Improve the json_responses_chunk_handler method to handle incomplete JSON chunks in stream. If a chunk does not end with }, it is considered incomplete and buffered until the next chunk arrives. This prevents JSON parsing errors and ensures all responses are processed correctly.

This PR is heavily inspired from #995 by @berkcaputcu, my implementation doesn't rely on exceptions and also includes additional specs.