airframe-parquet: Support writing generic records (e.g,. JSON, Map)

[xerial]

Parquet is a schema-first data format. To store JSON or Map in Parquet, we need to infer the schema of the data. Another approach is support setting a custom schema when creating a new Parquet writer.

Related:

• Spark supports Json schema inference: https://github.com/apache/spark/blob/master/sql/catalyst/src/main/scala/org/apache/spark/sql/catalyst/json/JsonInferSchema.scala
• Guess the schema of msgpack data Guess schema of msgpack data #516
• DataGuides https://www.vldb.org/conf/1997/P436.PDF (VLDB 1997)