Schema datasets

[Oxid15]

This one changes how data validation is handled in cascade. It introduces schemas in datasets.

Now pipeline blocks can explicitly define input schemas as pydantic models (optional dependency for this feature). If schema is defined in new SchemaModifier then it will implicitly insert ValidationWrapper before itself in the pipeline. When using self._dataset inside its own __getitem__ it will wrap itself in the validator which will perform validation of the input.
For example we have a dataset of annotated images:

class AnnotImage(pydantic.BaseModel):
    image: List[List[List[float]]]
    segments: List[List[int]]
    bboxes: List[Tuple[int, int, int, int]]

Base class for the pipeline will be:

class ImagesDataset(SchemaModifier):
    in_schema = AnnotImage

Let's define a dummy modifier:

class IDoNothing(ImagesDataset):
    def __getitem__(self, idx):
        item = self._dataset[idx]
        return item

Let's say we have a source of images and
hope that it will maintain our schema. The following
is the regular way we use modifier, but under the hood
it will automatically check that the output of the
datasource is AnnotImage

ds = SomeImageDatasource()
ds = IDoNothing(ds)