Fix sparse retrieval with filters returns results without any text-match

[tstadel]

Proposed changes:

• replace should operator of multi_match with must operator

How to reproduce?

• Run demo elasticsearch
• Hit postman with POST http://localhost:9200/document/_search and the following bodies

with should:

{
    "size": 10,
    "query": {
        "bool": {
            "should": [
                {"multi_match": {"query": "test", "type": "most_fields", "fields": ["content"]}}
            ],
            "filter": {"term": {"content_type": "text"}}
        }
    }
}

returns

...
"total": {
            "value": 3827,
            "relation": "eq"
        },
...

with must:

{
    "size": 10,
    "query": {
        "bool": {
            "must": [
                {"multi_match": {"query": "test", "type": "most_fields", "fields": ["content"]}}
            ],
            "filter": {"term": {"content_type": "text"}}
        }
    }
}

returns

...
"total": {
            "value": 66,
            "relation": "eq"
        },
...

Status (please check what you already did):

• First draft (up for discussions & feedback)
• Final code
• Added tests

[julian-risch]

LGTM! Linter isn't happy yet though: haystack/pipelines/utils.py:186:12: E4702: Iterated dict 'calculated_metrics' is being modified inside for loop body, iterate through a copy of it instead. (modified-iterating-dict)
A bonus would be to have a test case for the issue that is fixed by this PR.