feat: implement match_single for participant solution

[ali-sefidmouy]

No description provided.

[tavallaie]

1. Incorrect parent-child matching logic
   Using shared allele overlap (q_set & d_set) instead of Mendelian subset rule. This causes massive false positives — unrelated profiles with common alleles rank highly.

2. No proper Likelihood Ratio (CLR) calculation
   CLR set to number of "consistent" loci. Real forensic evaluation requires product of per-locus LRs using allele frequencies. Current ranking does not reflect true relationship strength.

3. No mutation handling
   True parent-child pairs with even one ±1 mutation are incorrectly penalized or excluded.

4. No candidate pre-filtering / indexing
   Full scan of ~500k profiles per query → extremely slow (likely times out in evaluation). Challenge requires efficient filtering for scalability.

5. Bidirectional logic not properly implemented
   Subset check must work both ways (query as child or query as parent), but current overlap test fails this.

6. Mutated_loci always 0
   Required output field not populated.

7. Inconclusive vs mismatch confusion
   True mismatches counted as "inconclusive" instead of exclusions.

Score impact: True parent rarely appears in top 10 due to false positives and lack of statistical weighting.