LIFT-2099: Add relevance scoring to wildcard search queries

[wes-maszk-rp]

Summary

• Adds MatchQuery.java - new Elasticsearch match query DSL class
• Modifies withWildCardPopulater() to generate hybrid queries with both wildcard (filter) and match (should)
• Updates BoolQuery to handle MatchQuery serialization

Problem

Multi-word searches like "pb promo" returned results in arbitrary order because wildcard queries don't contribute to Elasticsearch scoring - they're binary (match/no-match) and produce constant scores.

Solution: Hybrid Query Approach

Before (wildcard only):

{
  "bool": {
    "should": [
      {"wildcard": {"keywords": "*pb*"}},
      {"wildcard": {"keywords": "*promo*"}}
    ]
  }
}

After (hybrid filter + should):

{
  "bool": {
    "filter": [
      {"wildcard": {"keywords": "*pb*"}},
      {"wildcard": {"keywords": "*promo*"}}
    ],
    "should": [
      {"match": {"keywords": "pb"}},
      {"match": {"keywords": "promo"}}
    ]
  }
}

How It Works

Clause	Query Type	Purpose	Scoring
filter	Wildcard	Ensures substring exists	None (binary)
should	Match	Boosts exact token matches	TF-IDF/BM25

Why This Approach

1. Backwards Compatible - same documents match
2. Standard Elasticsearch Pattern - documented approach
3. No Reindexing - works with existing keywords field

Test plan

• CI passes
• Deploy to dev and test Portal search with multi-word queries
• Confirm "pb promo" returns exact matches ranked higher

[mason-chester-rp]

Multi-value wildcard queries: OR → AND semantic regression

For queries like sw(city,Chand,Atl), the loop in withWildCardPopulater adds each value's wildcard to filter and each match query to should:

{
  "bool": {
    "filter": [{"wildcard": {"city": "Chand*"}}, {"wildcard": {"city": "Atl*"}}],
    "should": [{"match": {"city": "Chand"}}, {"match": {"city": "Atl"}}]
  }
}

When filter clauses are present, minimum_should_match defaults to 0 — should is scoring-only. Matching is determined entirely by filter, and all filter clauses are ANDed. A city can't start with both "Chand" and "Atl", so this returns zero results.

The old behavior put wildcards in should with no filter, giving minimum_should_match=1 (OR semantics).

Fix: each wildcard+match pair needs its own nested BoolQuery, with those placed in an outer should:

{
  "bool": {
    "should": [
      {"bool": {"filter": [{"wildcard": {"city": "Chand*"}}], "should": [{"match": {"city": "Chand"}}]}},
      {"bool": {"filter": [{"wildcard": {"city": "Atl*"}}], "should": [{"match": {"city": "Atl"}}]}}
    ]
  }
}

This preserves OR semantics while still adding relevance scoring per term.

[mason-chester-rp]

Version mismatch: build.gradle was bumped to 0.3.6.17 but the changelog entry says 0.3.6.19.