Fix Qdrant tests on CI

[burnash]

Background

#998 makes qdrant working again but leaves a few things to be fixed. Qdrant is querying state in a wrong way (relies on order which is not there). Some tests are incorrect. Also we should test embedded qdrant.

Note we use following hack to go around ordering:

    def _create_point_no_vector(self, obj: Dict[str, Any], collection_name: str) -> None:
        """Inserts a point into a Qdrant collection without a vector.

        Args:
            obj (Dict[str, Any]): The arbitrary data to be inserted as payload.
            collection_name (str): The name of the collection to insert the point into.
        """
        # we want decreased ids because the point scroll functions orders by id ASC
        # so we want newest first
        id_ = 2**64 - int(precise_time() * 10**6)
        self.db_client.upsert(
            collection_name,
            points=[
                models.PointStruct(
                    id=id_,
                    payload=obj,
                    vector={},
                )
            ],
        )

Task

1. • decide if the keep the hack above or to enable ordering of records by created_at. ordering needs additional indexes, and indexes work only on server version of qdrant. so not sure it is worth the effort.
2. • add qdrant to all the tests where weaviate is added
3. • write additional tests that make sure we get the newest schema and right state when we have state that is not committed (no corresponding entry in the _dlt_loads collection)
4. • fix test_pipeline_replace which is loading same data twice
5. • enable local qdrant tests on CI (we have workflow with local tests)
6. • enable server qdrant tests

[sh-rp]

NOTE: To re-enable the failing tests remove them from SKIPPED_TESTS in docs/tools/prepare_examples_tests.py

[rudolfix]

@steinitzu #1545 is fixing some of issues above. IMO one still remain:

• test_pipeline_replace looks like loading same data twice so is not testing replace disposition
  pls do another PR for those two