[Datasets] Dataset.zip doesn't work if blocks have different types

[bveeramani]

What happened + What you expected to happen

Tried to zip two datasets, got:

(do_zip pid=52072) ValueError: Cannot zip <class 'ray.data._internal.pandas_block.PandasBlockAccessor'> with block of type <class 'pyarrow.lib.Table'>

Versions / Dependencies

Ray: 1012fbc

Reproduction script

import ray
import pandas as pd
df = pd.DataFrame({"spam": [0]})
ds1 = ray.data.from_pandas(df)
ds2 = ray.data.from_items([{"ham": [0]}])
ds1.zip(ds2)

Issue Severity

Medium: It is a significant difficulty but I can work around it.

[anyscalesam]

part of trend of aggregating Blocks of different type; for consideration as a larger project for ray210