Handle `None` in columns in defined in storage backends

[davidbp]

@jjlatval mentioned that providing None value in columns can cause unnexpected behaviours.

The following snippet can be executed

from docarray import Document, DocumentArray

DA_CONFIG = {
    "storage": "weaviate",
    "config": {
        "name": "Persisted",
        "host": "localhost",
        "port": 8080,
        "vector_cache_max_objects": 3000000,
        "columns": [("rank", "int"), ("year", "int")]
    }
}

da = DocumentArray(**DA_CONFIG)
docs = [Document(text="Some text here", rank=2, year=2020),
        Document(text="Another kind of text here", rank=1, year=2022),
        Document(text="This document is bad", rank=2, year=3)]

with da:
    da.clear()
    da.extend(docs)

da[:,'tags__rank']

But when we have a None value then we can't index to a Document containing a None in a column that was specified with a certain datatype.

from docarray import Document, DocumentArray

DA_CONFIG = {
    "storage": "weaviate",
    "config": {
        "name": "Persisted",
        "host": "localhost",
        "port": 8080,
        "vector_cache_max_objects": 3000000,
        "columns": [("rank", "int"), ("year", "int")]
    }
}

da = DocumentArray(**DA_CONFIG)
docs = [Document(text="Some text here", rank=2, year=2020),
        Document(text="Another kind of text here", rank=1, year=2022),
        Document(text="This document is bad", rank=None, year=3)]

with da:
    da.clear()
    da.extend(docs)

# this will break
da[:,'tags__rank']