Provide structured validation error in Python bindings

[Stranger6667]

At the moment, it is not possible to safely define a custom exception class via PyO3 that will expose some Rust struct fields as Python-side instance attributes.

The solution is to add an intermediate Python module that will wrap validate, catch the underlying exception, and re-raise it via a different class.

from . import jsonschema_rs as rs


is_valid = rs.is_valid
JSONSchema = rs.JSONSchema
Draft4 = rs.Draft4
Draft6 = rs.Draft6
Draft7 = rs.Draft7
__build__ = rs.__build__


class ValidationError(ValueError):

    def __init__(self, message, instance_path, instance_path_pointer):
        super().__init__(message)
        self.instance_path = instance_path
        self.instance_path_pointer = instance_path_pointer


ValidationError.__doc__ = rs.ValidationError.__doc__


def validate(schema, instance, draft=None, with_meta_schemas=False):
    try:
        return rs.validate(schema, instance, draft, with_meta_schemas)
    except rs.ValidationError as exc:
        raise ValidationError(
            message=exc.args[0], 
            instance_path=exc.args[1], 
            instance_path_pointer=exc.args[2]
        ) from None

validate.__doc__ = rs.validate.__doc__

I am not sure about attributes - maybe it will be better to follow the same structure as in the ValidationError struct - separate class for instance_path that can be converted to a string or a vector of strings.

This approach implies some overhead, but preliminary tests did not indicate any significant slowdown.

As an extra benefit - it will be possible to specify type annotations for functions created on the python-level, so maybe it worth wrapping is_valid as well.

The snippet above is almost sufficient. Other things:

• Update raise_on_error in Python bindings code, so it passes a tuple of needed attributes to the bindings-level ValidationError;
• Update the build system - the resulting Python package should include this code.

[Stranger6667]

When PyO3/pyo3#1591 gets merged & released, the abovementioned intermediate Python level won't be needed, so it might be better to wait a bit.