Bug: delete_where fail sanity check when a service is initialized with statment

[noamsto]

Description

This is somewhat unexpected behavior,
Since delete is not using the base statement from the service, sanity checks may fail.

For example,
I have a service which limit the access by my tenant_id:

async def provide_document_access_received_service(
    request: "Request[UserAccount, Token, Any]", db_session: AsyncSession
) -> AsyncGenerator[ResourceAccessService, None]:
    """Provider for resource access service."""
    async with ResourceAccessService.new(
        statement=select(ResourceAccess)
        .where(ResourceAccess.tenant_id.in_(request.user.accessible_tenant_ids))
        .filter(ResourceAccess.resource.has(Resource.resource_type == ResourceType.DOCUMENT)),
        load=selectin_polymorphic(Resource, [Document]),
        session=db_session,
    ) as service:
        yield service

Now let's say I try to delete a resource which is not in my tenant, I expect to get NotFoundError, but in this case I'll get Sanity Check Failed exception.
Unless I'll provide the same filters I provided with the statement.

A possible solution, which I'm thinking to do and create a PR (if I'll have time)
Is have the new method get filters and build the statement internally, this way these "global" filters can be applied to the delete_where and the behavior will be more predictable.

URL to code causing the issue

No response

MCVE

No response

Steps to reproduce

No response

Screenshots

Logs

No response

Package Version

v0.23.1

Platform

• Linux
• Mac
• Windows
• Other (Please specify in the description above)