v1.0.1 — Python parity fixes + Turkish-locale bug fix

[1.0.1] - 2026-07-22

Not backward compatible with 1.0.0. The previous release's algorithms were
silently wrong for the majority of types — this release corrects them to match
the Python (mockjutsu) ground truth exactly. Any value previously generated
by 1.0.0 for an affected type (checksum, format, or structure) will differ
under 1.0.1. 1.0.0 is deprecated and should not be used.

Fixed

• 273 mismatches found and corrected across all 39 generator files, via a
  full audit comparing every shared type against the Python engine's real
  output (not just structural/exception checks).
• ean13Check() (shared by EAN-13, EAN-8, UPC-A, GS1-128) — checksum weight
  assignment was only coincidentally correct for even-length payloads
  (EAN-13); odd-length payloads (EAN-8, UPC-A, GS1-128) produced invalid
  check digits. Verified against official GS1 test vectors.
• EcommerceGen.dhlTracking() — Luhn checksum was computed over the "JD"
  prefix letters instead of the numeric body only.
• Turkish-locale JVM default String.toUpperCase()/toLowerCase() producing
  dotted/dotless İ instead of ASCII I in card-network and IBAN generation.
• OidcGen — real P-256 ECDSA (ES256) signing implemented from scratch
  (previously not cryptographically valid), verified against independent
  signature verification.
• Fido2Gen — real CBOR-encoded WebAuthn/FIDO2 attestation objects
  (previously placeholder structure).
• Numerous str(dict)-vs-json.dumps() output format mismatches (IoT and
  commerce types returning Python repr syntax instead of JSON).
• Wallet/crypto address derivation, BIP-39 mnemonics, and ~30 additional
  algorithm-level bugs from the original audit pass (see project memory for
  full list).
• Turkish-locale JVM default String.format("%f"...) (no explicit Locale)
  producing comma instead of period as the decimal separator — found in
  TleGen (epoch day), EcommerceGen (rating), FinancialGen (balance),
  and LocationGen (latitude/longitude). Caught by running the full existing
  JUnit suite (4183 tests) against the fix, which the original 273-mismatch
  audit had not exercised.
• Removed sanctions_hit_masked — a dead type registered in the JMeter
  dispatcher with no backing implementation and no Python counterpart.

Changed

• TelecomGen.imei() and BarcodeGen.ean13() signatures adjusted as part of
  the correctness fixes; internal unit tests updated to match.
• Updated ~35 pre-existing unit test assertions that encoded the old (Java-only,
  pre-1.0.1) output format/casing/values instead of the real Python-parity
  behavior — verified each individually against Python ground truth before
  changing (casing of card type/status/category, IoT str(dict) vs JSON
  quoting, UK NIN/PAYE/German RVN/HRB spacing, tx_hash/signature/
  timestamp_iso/swift_mt103 format assumptions, and others).