[BUG] Spark execution engine can't take empty pandas dataframes

[goodwanghan]

Minimal Code To Reproduce

df = pd.DataFrame([],columns=["a","b"])

with FugueSQLWorkflow("spark") as dag:
    dag.df(df).show()

Describe the bug

ValueError: can not infer schema from empty dataset

Expected behavior
Although this is a problem of Spark, we should fix it through Fugue level, also we need to make sure all engines can take empty pandas dataframes

Environment (please complete the following information):

• Backend: pandas/dask/ray?
• Backend version:
• Python version:
• OS: linux/windows