Improving documentation about es.read.fields.include, and fixing a related bug

[masseyke]

The documentation for es.read.fields.include left room for confusion. Also the bevaior was different between spark 1,
spark 2, and spark 3 -- setting "es.read.fields.include" to part of a hierarchy caused a NullPointerException in spark 1
and spark 2.
Closes #1784

[masseyke]

This is confusing, and easiest for me to think about using concrete examples. I added them in the integration tests in this PR, but it's probably easier to see without code. Using the schema from #1784:
root

 |-- features: struct (nullable = true)
 |    |-- hashtags: array (nullable = true)
 |    |    |-- element: struct (containsNull = true)
 |    |    |    |-- count: long (nullable = true)
 |    |    |    |-- text: string (nullable = true)

Say I have this data in elasticsearch:

features:
    hashtags:
        [
            {"text": "foo"},
            {"count":"bar"}
        ] 

If I leave "es.read.fields.include" unset or if I set it to "features.hashtags.*" I get

features:
    hashtags:
        [
            {"text": "foo"},
            {"count":"bar"}
        ] 

If I set "es.read.fields.include" to "features.hashtags" I get

features:
    hashtags:
        [
            {},
            {}
        ] 

If I set "es.read.fields.include" to "features.hashtags.text" I get

features:
    hashtags:
        [
            {"text": "foo"},
            {}
        ] 

[jbaiera]

LGTM, thanks for clarifying these.