[ES|QL] Provide the ability to have `SORT` ignore non-alphanumeric characters

[BenB196]

Description

Today, when using SORT with strings (keywords) it takes into account non-alphanumeric (punctual) characters (ex: ( or ") and sorts these first.

Example:

ROW example = ["abc", "def", "\"ghij", "(x)yz"]
| MV_EXPAND example
| SORT example

Gets you:

{
  "rawResponse": {
    "columns": [
      {
        "name": "example",
        "type": "keyword"
      }
    ],
    "values": [
      [
        "\"ghij"
      ],
      [
        "(x)yz"
      ],
      [
        "abc"
      ],
      [
        "def"
      ]
    ]
  }
}

In some other systems (ex: Linux ls command), it's possible to ignore these characters and only sort on the alphanumeric values.

It'd be nice to have the ability to configure this within ES|QL.

[elasticsearchmachine]

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

[WeiTanZzz]

My solution is to populate a new field named pureString and sort by this new field. It could be not performant when meeting a large data set.

POST _query?format=txt
{
  "query":"""
  ROW example = ["abc", "def", "\"ghij", "(x)yz"]
  | MV_EXPAND example
  | EVAL pureString = REPLACE(example, "[^a-zA-Z0-9]", "")
  | SORT pureString
  """
}

    example    |  pureString   
---------------+---------------
abc            |abc            
def            |def            
"ghij          |ghij           
(x)yz          |xyz