geo_point类型转换失败

# 环境/配置
* es: 6.3.2
* "cleanup": true

# 情况一：三个字段，类型分别为date, geo_point, date时

```sql
select
current_date as date,
'23.0391693115234,113.137992858887' as location,
'2012-12-12' as name
```

```json
"column": [
{
"name": "date",
"type": "date"
},
{
"name": "location",
"type": "geo_point"
},
{
"name": "name",
"type": "date"
}
]
```

写入成功后，mapping没问题
```json
{
"ar_test_geo": {
"aliases": {},
"mappings": {
"_doc": {
"properties": {
"date": {
"type": "date",
"format": "strict_date_optional_time||epoch_millis||yyyy-MM-dd HH:mm:ss||yyyy-MM-dd"
},
"location": {
"type": "geo_point"
},
"name": {
"type": "date",
"format": "strict_date_optional_time||epoch_millis||yyyy-MM-dd HH:mm:ss||yyyy-MM-dd"
}
}
}
}
}
```

# 情况二：三个字段，类型分别为date, geo_point, string时

```sql
select
current_date as date,
'23.0391693115234,113.137992858887' as location,
'aa' as name
```

```json
"column": [
{
"name": "date",
"type": "date"
},
{
"name": "location",
"type": "geo_point"
},
{
"name": "name",
"type": "string"
}
]
```

写入成功后，geo_point转成了text，为啥？
```json
{
"ar_test_geo": {
"aliases": {},
"mappings": {
"_doc": {
"properties": {
"date": {
"type": "date"
},
"location": {
"type": "text",
"fields": {
"keyword": {
"type": "keyword",
"ignore_above": 256
}
}
},
"name": {
"type": "text",
"fields": {
"keyword": {
"type": "keyword",
"ignore_above": 256
}
}
}
}
}
}
}
```

# 暂时解决方法
只能通过writer的settings配置来指定mapping了，这个就和column配置冗余了，同时需要管理两份元数据。

Reproduce both SQL cases with Elasticsearch 6.3.2, comparing the column configuration with the writer settings that define the mapping. Inspect the writer's mapping-generation entry point and determine why a geo_point becomes text when followed by a string column. Done means the generated mapping preserves geo_point without requiring duplicate mapping metadata.

Written by the indexing model from the issue text.