## What changes were proposed in this pull request?

Originally this PR was based on #14491 but I realised that fixing examples are more sensible rather than comments.

This PR fixes three things below:

 - Fix two wrong examples in `structured-streaming-programming-guide.md`. Loading via `read.load(..)` without `as` will be `Dataset<Row>` not `Dataset<String>` in Java.

- Fix indentation across `structured-streaming-programming-guide.md`. Python has 4 spaces and Scala and Java have double spaces. These are inconsistent across the examples.

- Fix `StructuredNetworkWordCountWindowed` and  `StructuredNetworkWordCount` in Java and Scala to initially load `DataFrame` and `Dataset<Row>` to be consistent with the comments and some examples in `structured-streaming-programming-guide.md` and to match Scala and Java to Python one (Python one loads it as `DataFrame` initially).

## How was this patch tested?

N/A

Closes https://github.com/apache/spark/pull/14491

Author: hyukjinkwon <gurwls223@gmail.com>
Author: Ganesh Chand <ganeshchand@Ganeshs-MacBook-Pro-2.local>

Closes #14564 from HyukjinKwon/SPARK-16886.