Support `sameElement` filter semantics in Vespa grouping

[schandrasekhar]

Is your feature request related to a problem? Please describe.
Vespa grouping language does not support sameElement filter semantics. Example shown below
Vespa schema

struct string_pair {
            field pair_key type string {}
            field pair_val type string {}
        }
        field meta_data type array<string_pair> {
            indexing: summary
            struct-field pair_key {
                indexing: attribute
                attribute: fast-search
            }
            struct-field pair_val {
                indexing: attribute
                attribute: fast-search
            }
        }

Sample document

{
"meta_data": [
                        {
                            "pair_key": "at_999_9689",
                            "pair_val": "33194"
                        },
                        {
                            "pair_key": "at_999_146785",
                            "pair_val": "476289"
                        }
]
}

For this yql query

"select * from sources stuff where true | all(group('1') filter(regex('(at_999_9689)', meta_data.pair_key) and regex('(476289)', meta_data.pair_val)) each(output(count())))"

It returns count()  as 1 even though the meta_data.pair_key "at_999_9689" does not have pair_value "476289". It looks like the grouping is not following the sameElement filter semantics. Can you please provide this support?

Describe the solution you'd like
Ran a parallel query with sameElement filter and got the document count.

Describe alternatives you've considered
A clear and concise description of any alternative solutions or features you've considered.

Additional context
Add any other context or screenshots about the feature request here.