Use nullable inferred schema in function apply

[HyukjinKwon]

This PR proposes to use nullable schema in when to apply a function. We take some to infer the schema. Usually types are matched but null or NaN is sparse often. This behaviour is batch with Spark's JSON schema inference as well.

from pyspark.sql import SparkSession
import databricks.koalas as ks
import numpy as np

spark = SparkSession.builder.getOrCreate()
data = list()
for i in range(1, 10000):
  data.append((str(i % 100), np.nan if i % 9999 == 0 else float(i),))

sdf = spark.createDataFrame(data, 'a string, b float').repartition(10)
kdf = sdf.to_koalas()

def f(df):
  return df

df = kdf\
  .groupby('a')\
  .apply(f)

df.to_spark().printSchema()
df[df['b'] == -1]

Before:

root
 |-- a: string (nullable = false)
 |-- b: float (nullable = false)

java.lang.IllegalStateException: Value at index is null
...

After:

root
 |-- a: string (nullable = true)
 |-- b: float (nullable = true)

Empty DataFrame
Columns: [a, b]
Index: []

Resolves #1885

[ueshin]

LGTM.

[ueshin]

Thanks! merging.