Fix BM25 score for PhraseDocIterator (#2404)

### What problem does this PR solve?

Update IDF formula for PhraseDocIterator
reference:
https://lucene.apache.org/core/10_1_0/core/org/apache/lucene/search/similarities/BM25Similarity.html#idfExplain(org.apache.lucene.search.CollectionStatistics,org.apache.lucene.search.TermStatistics[])

Issue link:#1320

### Type of change

- [x] Bug Fix (non-breaking change which fixes an issue)
- [x] Refactoring
- [x] Test cases

src/storage/invertedindex/search/phrase_doc_iterator.cpp

@@ -42,10 +42,14 @@ void PhraseDocIterator::InitBM25Info(UniquePtr<FullTextColumnLengthReader> &&col
     constexpr float b = 0.75F;
 
     column_length_reader_ = std::move(column_length_reader);
-    u64 total_df = column_length_reader_->GetTotalDF();
-    float avg_column_len = column_length_reader_->GetAvgColumnLength();
-    float smooth_idf = std::log1p((total_df - estimate_doc_freq_ + 0.5F) / (estimate_doc_freq_ + 0.5F));
-    bm25_common_score_ = weight_ * smooth_idf * (k1 + 1.0F);
+    const u64 total_df = column_length_reader_->GetTotalDF();
+    const float avg_column_len = column_length_reader_->GetAvgColumnLength();
+    float total_idf = 0.0f;
+    for (const auto &iter : pos_iters_) {
+        const auto doc_freq = iter->GetDocFreq();
+        total_idf += std::log1p((total_df - doc_freq + 0.5F) / (doc_freq + 0.5F));
+    }
+    bm25_common_score_ = weight_ * total_idf * (k1 + 1.0F);
     bm25_score_upper_bound_ = bm25_common_score_ / (1.0F + k1 * b / avg_column_len);
     f1 = k1 * (1.0F - b);
     f2 = k1 * b / avg_column_len;

test/sql/dql/fulltext/fulltext.slt

@@ -36,26 +36,26 @@ Anarchism 30-APR-2012 03:25:17.000 0 22.299635
 query TTIR rowsort
 SELECT doctitle, docdate, ROW_ID(), SCORE() FROM sqllogic_test_enwiki SEARCH MATCH TEXT ('body^5', '"social customs"', 'topn=3;block_max=compare') USING INDEXES ('ft_index');
 ----
-Anarchism 30-APR-2012 03:25:17.000 6 20.753590
+Anarchism 30-APR-2012 03:25:17.000 6 27.133215
 
 # only phrase
 query TTIR rowsort
 SELECT doctitle, docdate, ROW_ID(), SCORE() FROM sqllogic_test_enwiki SEARCH MATCH TEXT ('body^5', '"social customs"', 'topn=3;block_max=compare');
 ----
-Anarchism 30-APR-2012 03:25:17.000 6 20.753590
+Anarchism 30-APR-2012 03:25:17.000 6 27.133215
 
 # phrase and term
 query TTIR rowsort
 SELECT doctitle, docdate, ROW_ID(), SCORE() FROM sqllogic_test_enwiki SEARCH MATCH TEXT ('body^5', '"social customs" harmful', 'topn=3');
 ----
 Anarchism 30-APR-2012 03:25:17.000 0 22.299635
-Anarchism 30-APR-2012 03:25:17.000 6 20.753590
+Anarchism 30-APR-2012 03:25:17.000 6 27.133215
 
 # phrase and term
 query TTIR rowsort
-SELECT doctitle, docdate, ROW_ID(), SCORE() FROM sqllogic_test_enwiki SEARCH MATCH TEXT ('body^5', '"social customs" harmful', 'topn=3;threshold=21.5');
+SELECT doctitle, docdate, ROW_ID(), SCORE() FROM sqllogic_test_enwiki SEARCH MATCH TEXT ('body^5', '"social customs" harmful', 'topn=3;threshold=25');
 ----
-Anarchism 30-APR-2012 03:25:17.000 0 22.299635
+Anarchism 30-APR-2012 03:25:17.000 6 27.133215
 
 # copy data from csv file
 query I