feat(real-world-vfc-slice): add labeled OpenSSL/Firefox/Chromium/curl pair corpus

Harvest 176 isolated vuln-vs-fix functions (109 CVEs, 352 training rows)
into benchmarks/pairs/vfc with provenance and a JSONL training manifest.
Score the slice as an honesty dashboard (ousast pairs --slice vfc);
keep the pair gate on the local fixtures only.

benchmarks/pairs/vfc/openssl-cve-2022-4304-rsa_blinding_invert/fixed.c

@@ -0,0 +1,25 @@
+/* Provenance: openssl/openssl rsa_blinding_invert (fixed).
+ * repo: openssl/openssl
+ * commit: a00d757d9ca212994625d1a02c81cc5edd27e13b
+ * parent: 908eaceb62624f5b5c505b286d904bd3a4e8a64a
+ * commit_url: https://github.com/openssl/openssl/commit/a00d757d9ca212994625d1a02c81cc5edd27e13b
+ * cve: CVE-2022-4304
+ * license: Apache-2.0
+ * function: rsa_blinding_invert
+ * relpath: crypto/rsa/rsa_ossl.c
+ */
+
+static int rsa_blinding_invert(BN_BLINDING *b, BIGNUM *f, BIGNUM *unblind,
+                               BN_CTX *ctx)
+{
+    /*
+     * For local blinding, unblind is set to NULL, and BN_BLINDING_invert_ex
+     * will use the unblinding factor stored in BN_BLINDING. If BN_BLINDING
+     * is shared between threads, unblind must be non-null:
+     * BN_BLINDING_invert_ex will then use the local unblinding factor, and
+     * will only read the modulus from BN_BLINDING. In both cases it's safe
+     * to access the blinding without a lock.
+     */
+    BN_set_flags(f, BN_FLG_CONSTTIME);
+    return BN_BLINDING_invert_ex(f, unblind, b, ctx);
+}