ESQL: Load from _source if a field was ignore_above-ed

[nik9000]

Description

Elasticsearch's ignore_above feature skips indexing a keyword field if it is longer than some number of characters. That's useful to preserve lucene's mind, but we lose the ability to search that field. It'd be good if we could load the field form _source in that case.

Example:

curl -uelastic:password -XDELETE localhost:9200/long_field
curl -uelastic:password -XPUT -HContent-Type:application/json localhost:9200/long_field?pretty -d'{
  "mappings": {
    "properties": {
      "url": {
        "properties": {
          "query": {
            "type": "keyword",
            "ignore_above": 10
          }
        }
      }
    }
  },
  "settings": {}
}'

curl -uelastic:password -XPOST -HContent-Type:application/json localhost:9200/long_field/_doc?refresh -d'{
  "url": {
    "query": "short"
  }
}'
curl -uelastic:password -XPOST -HContent-Type:application/json localhost:9200/long_field/_doc?refresh -d'{
  "url": {
    "query": "words words words"
  }
}'
curl -uelastic:password -XPOST -HContent-Type:application/json localhost:9200/_query?pretty -d'{
  "query": "FROM long_field"
}'

Makes:

["short"],
[null]

But after this change it'd make:

["short"],
["words words words"]

These ignored fields are indexed and have a special doc value:

                context.doc().add(new SortedSetDocValuesField(NAME, new BytesRef(ignoredField)));
                context.doc().add(new StringField(NAME, ignoredField, Field.Store.NO));

This is generally cheap to maintain because there aren't that many unique field values and it powers the missing query. We could use the doc values to load these fields from _source. Pushed down queries are much harder - they'd need to double check the condition. I think. Something like that.

[elasticsearchmachine]

Pinging @elastic/es-analytical-engine (Team:Analytics)