[BUG] [Analyzer] [Token Filters] pattern_capture loses diacritic sign from beginning and the end of the word

[Roboteus]

Describe the bug

This is something specific to OpenSearch because I checked and looks like ElasticSearch has no issues. Unfortunatelly I compared to old version...

PUT someindex
{
	"settings": {
		"analysis": {
			"char_filter": {
				"text_number_cleaner": {
					"type": "pattern_replace",
					"pattern": "[^\\p{L}0-9-]",
					"replacement": " "
				}
			},
			"filter": {
				"capture_words": {
					"type": "pattern_capture",
					"patterns": [
						"(\\b\\p{L}+\\b)"
					]
				}
			},
			"analyzer": {
				"my_analyzer": {
					"type": "custom",
					"char_filter": [
						"text_number_cleaner"
					],
					"tokenizer": "keyword",
					"filter": [
						"lowercase",
						"capture_words"
					]
				}
			}
		}
	}
}

Now test results of the analyzer:

GET someindex
{
  "analyzer" : "my_analyzer",
  "text" : "źryj-jótro-kupę"
}

Related component

Indexing

To Reproduce

1. Create index with custom analyzer described in this ticket
2. Test the analyzer with words containing diacritic signs
3. Check results

Stated result

Tokens created by pattern_capture are: [ ryj , jótro , kup ]
Looks like words followed or ended by diacritic sign are trimmed by this sign

Expected behavior

[ źryj , jótro , kupę ]

Additional Details

Plugins
Please list all plugins currently enabled.

Screenshots

Host/Environment (please complete the following information):

• Version [2.18]

Additional context
Add any other context about the problem here.