[SPARK-17388] [SQL] Support for inferring type date/timestamp/decimal for partition column

## What changes were proposed in this pull request?

Currently, Spark only supports to infer `IntegerType`, `LongType`, `DoubleType` and `StringType`.

`DecimalType` is being tried but it seems it never infers type as `DecimalType` as `DoubleType` is being tried first. Also, it seems `DateType` and `TimestampType` could be inferred.

As far as I know, it is pretty common to use both for a partition column.

This PR fixes the incorrect `DecimalType` try and also adds the support for both `DateType` and `TimestampType` for inferring partition column type.

## How was this patch tested?

Unit tests in `ParquetPartitionDiscoverySuite`.

Author: hyukjinkwon <gurwls223@gmail.com>

Closes #14947 from HyukjinKwon/SPARK-17388.