Make Spark Serving Easier to Use

@mhamilton723 is already working on this.

Since Jul 18, 2018.

- [ ] Add option to fail when it cannot parse the input
- [x] Add tests on ADB
- [x] Add syntactic sugars for adding input parsers
- [x] Add ability to reply with other types
- [ ] Fix offsets commited out of order exception

Repro:

```python
import mmlspark
from pyspark.sql.functions import length, col, lit, from_json
from pyspark.sql.types import *

df = spark.readStream.server() \
.address("localhost", 8888, "my_api") \
.load()

server = df.withColumn("reply", length(col("value"))) \
.withColumn("replyString", col("reply").cast(StringType()))\
.writeStream \
.server() \
.option("name", "my_api") \
.queryName("my_query") \
.option("replyCol", "replyString") \
.option("checkpointLocation", "checkpoints") \
.start()

import requests

r = requests.post(data="{foo:1}", url="http://localhost:8888/my_api")
print(r.text)