Introduce IPRangeBucket type for IpRange aggregation

[russcam]

The current RangeBucket implementation ignores "from" and "to" for an ip_range aggregation, by looking at the JSON token type when deserializing, and only deserializing when it is a number:

elasticsearch-net/src/Nest/Aggregations/AggregateJsonConverter.cs

Lines 576 to 587 in 28f3774

	case Parser.From:
	reader.Read();
	if (reader.ValueType == typeof(double))
	fromDouble = (double)reader.Value;
	reader.Read();
	break;
	case Parser.To:
	reader.Read();
	if (reader.ValueType == typeof(double))
	toDouble = (double)reader.Value;
	reader.Read();
	break;

This ignores "from" and "to" string values for ip_range aggregation, an example response being

{
  "took" : 39,
  "timed_out" : false,
  "_shards" : {
    "total" : 2,
    "successful" : 2,
    "skipped" : 0,
    "failed" : 0
  },
  "hits" : {
    "total" : 1100,
    "max_score" : 0.0,
    "hits" : [ ]
  },
  "aggregations" : {
    "ip_range#ip_ranges" : {
      "buckets" : [
        {
          "key" : "*-127.0.0.1",
          "to" : "127.0.0.1",
          "doc_count" : 52
        },
        {
          "key" : "127.0.0.1-*",
          "from" : "127.0.0.1",
          "doc_count" : 48
        }
      ]
    }
  }
}

Introduce an IpRangeBucket type or make RangeBucket a generic type, i.e.

public class RangeBucket<TRangeValue> : BucketBase, IBucket
{
	public RangeBucket(IReadOnlyDictionary<string, IAggregate> dict) : base(dict) { }

	public long DocCount { get; set; }
	public TRangeValue From { get; set; }
	public string FromAsString { get; set; }

	public string Key { get; set; }
	public TRangeValue To { get; set; }
	public string ToAsString { get; set; }
}

An IpRangeBucket type is probably better here, since FromAsString and ToAsString are not properties of the response.

[russcam]

This is now merged into 7.x in #3647