Support infer batch

[continue-revolution]

This issue tracks infer_batch support

maybe can also resolve #103

[continue-revolution]

Also deal with this:

Example: one table "dataset" has columns like this

{
    "dataset": [
        {"name": "masked_image_path", "dtype": str, "is_primary_key": True},
        {"name": "original_image_path", "refers_to": ("character", "path"), "dtype": str},
        {"name": "tag_path", "dtype": str},
        {"name": "tag", "dtype": str}
    ]
}

Inference service like this

self.engines = [
    {
        "service": self.segmentor,
        "input_col": ("character.character_path",),
        "output_col": ("dataset.masked_image_path", "dataset.original_image_path")
    },
    {
        "service": self.tagger,
        "input_col": ("dataset.masked_image_path",),
        "output_col": ("dataset.tag_path", "dataset.tag")
    }
]

This cannot pass https://github.com/ddkang/aidb/blob/main/aidb/config/config.py#L152 We should allow output columns that are copied from input columns to be binded to 2 inference services