[Tests] Adapt QueryStringQueryBuilderTests expectations

[cbuescher]

With the introduction of field aliases, the toQuery() method now can also return
BlendedTermQuery. Added this to the list of expected query classes.

Closes #32234

[elasticmachine]

Pinging @elastic/es-search-aggs

[cbuescher]

@jimczi @jtibshirani the query string query that gets translated to a Lucene BlendedTermQuery in the previously failing test is:

{
  "query_string" : {
    "query" : "sMPjDLuo ",
    "fields" : [
      "mapped_string^0.4345442",
      "mapped_string_alias^1.0"
    ],
    "type" : "cross_fields",
    "tie_breaker" : 0.5032719,
    "default_operator" : "or",
    "max_determinized_states" : 10000,
    "allow_leading_wildcard" : true,
    "enable_position_increments" : true,
    "fuzziness" : "1",
    "fuzzy_prefix_length" : 0,
    "fuzzy_max_expansions" : 50,
    "phrase_slop" : 3,
    "rewrite" : "scoring_boolean",
    "minimum_should_match" : "-25%",
    "quote_field_suffix" : "Qcu",
    "escape" : false,
    "auto_generate_synonyms_phrase_query" : true,
    "fuzzy_transpositions" : true,
    "boost" : 1.0,
    "_name" : "paAge15"
  }
}

It includes the newly introduced alias fields, so I assume the behaviour is correct, but can you re-check that adding this to the expected list of query classes is okay?

[cbuescher]

On a second note, I also wouldn't be against removing these assertions altogether. In my opinion this makes the test more brittle and doesn't achieve much in terms of assertions, since the list of accepted classes is quite long already. Let me know if you also would approve if we remove this altogether.

[jimczi]

+1 to remove these assertions, the query_string query can build all sort of queries so I don't think we should track the full list here. This will also allow more flexibility in the random query generator.

[cbuescher]

@jimczi thanks for your comment, I pushed an update removing the assertions altogether

[jimczi]

LGTM

[jtibshirani]

Thanks @cbuescher for fixing this!