Fix wrong use of currentFieldName outside of a parsing loop

[obourgain]

Parsing a multipercolate request copies the content of the JSON to a Map and then iterates on the entrySet, but for the fields ignore_unavailable, allow_no_indices and expand_wildcards it is still using the currentFieldName variable and not entry.getKey().
This cause the parsing to fail if the last fieldName of the JSON was one of these fields.

example :

curl -XPOST "http://localhost:9200/twitter/_mpercolate" -d'
{"percolate":{"index":"the_index","type":"the_type", "realtime": true,"expand_wildcards":"open"}}

will fail with error :

{
   "error": "ElasticsearchIllegalArgumentException[No valid expand wildcard value [true]]",
   "status": 400
}

This PR fixes the parsing and also extracts the parsing to a map to a dedicated method to avoid having the parsing vars wrongly reused in the future.

[jpountz]

The fix looks good. Thanks for extracting the parsing logic to another method in order to make sure we don't make the same mistake again if we add a new parameter! Will merge soon.