Support full parameter set for Ollama integrations

[emso-c]

Is your feature request related to a problem? Please describe.

Ollama integrations do not support the full capabilities of the generation methods. Namely OllamaGenerator.generate and OllamaChatGenerator.chat methods. I was trying to add the keep_alive option to the generation using the OllamaGenerator. But when I inspected the code, I've seen that the method only passed down the model, prompt, stream and options parameters, even though the Ollama client is capable of utilizing more:

@overload
  def generate(
    self,
    model: str = '',
    prompt: str = '',
    suffix: str = '',
    system: str = '',
    template: str = '',
    context: Optional[Sequence[int]] = None,
    stream: Literal[True] = True,
    raw: bool = False,
    format: Literal['', 'json'] = '',
    images: Optional[Sequence[AnyStr]] = None,
    options: Optional[Options] = None,
    keep_alive: Optional[Union[float, str]] = None,
  ) -> Iterator[Mapping[str, Any]]: ...

Describe the solution you'd like
Adding all the top level parameters that ollama client is able to use to the OllamaGenerator and OllamaChatGenerator constructors respectively.

Describe alternatives you've considered
Using kwargs to pass down the parameters manually:

class OllamaGenerator:
    def __init__(
        ...
        run_kwargs: Optional[Dict[str, Any]] = None # added line
    ) 
    ...
    def run(
        ...
        response = self._client.generate(model=self.model,
                                         prompt=prompt,
                                         stream=stream,
                                         options=generation_kwargs,
                                         **self.run_kwargs) # added line
        ...

# usage
generator = OllamaGenerator(
    run_kwargs={"keep_alive": "5m"},
)

Additional context

[anakin87]

Some of the available parameters are already supported.

For now, I would only add support for keep_alive, which seems important.

Feel free to create a PR if you want, otherwise we will take care of it in the future.

[anakin87]

Support for keep_alive was added in #1131.

Support for images will be added when we focus more on enabling multimodal use cases throughout the framework.