feat: add benchmark-programs — 58 telecom COBOL test programs

作为子目录纳入系统，与核心测试管道协同

Co-Authored-By: Claude <noreply@anthropic.com>

benchmark-programs/20-matching-MN-to-MxN/README.md

@@ -0,0 +1,37 @@
+# 20-matching-MN-to-MxN
+
+## 电信业务场景
+
+CDR详细清单M:N→M×N。多个合同×多个通话明细的笛卡尔积组合，输出完整的CDR详细清单(M×N件)。
+
+**M:N matching — Cartesian product output (M x N records per matching key).**
+
+Tests the scenario where both input files may contain duplicate keys and
+every combination of matching master and detail records must be output
+(the full Cartesian product).
+
+## Test data
+
+| File        | Records | Keys                                     |
+|-------------|---------|------------------------------------------|
+| master.dat  | 5       | KEY00001 (x2), KEY00002 (x1), KEY00003 (x2) |
+| detail.dat  | 7       | KEY00001 (x3), KEY00003 (x2), KEY00004 (x2) |
+
+- **KEY00001:** 2 masters x 3 details = 6 output records
+- **KEY00003:** 2 masters x 2 details = 4 output records
+- **Total output:** 10 records
+- **Unmatched:** KEY00002 (master only), KEY00004 (detail only)
+
+## Algorithm
+
+1. Read both files synchronously (both assumed sorted by key).
+2. When keys match: read all master records for that key into a table,
+   read all detail records for that key into a table, then write M x N
+   output records covering every combination.
+3. Unmatched groups are skipped via the standard key-comparison logic.
+
+## Expected output
+
+- **output.dat:** 10 records x 45 bytes = 450 bytes
+- Program displays: `20-matching-MN-to-MxN: PASS`
+- Script displays: `20-matching-MN-to-MxN: ALL TESTS PASSED`