Generator approach in harvesting

For larger harvests, it makes sense to push item to the queue iteratively, instead of waiting for the whole result set.

This can be achieved quite easily by transforming the endpoint.harvest method to yield items instead of return them. Also the cql_filter function must be adapted. The stringify function could look like this:

def stringify(
    result: Generator[dict], mode: str = "item", extract_property: Optional[str] = None
):
    if mode == "item":
        yield from (json.dumps(item, default=str) for item in result)
    elif mode == "property":
        yield from (item["properties"][extract_property] for item in result)

The harvester main looks quite unchanged, only the client.lpush(harvest_config["queue"], *encoded) must be done in a for-loop.