Update CTranslate2 to fix CUBLAS_STATUS_NOT_SUPPORTED on RTX 50XX / Blackwell

I am encountering the cuBLAS failed with status CUBLAS_STATUS_NOT_SUPPORTED error when attempting to run the standalone executable. This prevents the use of the default int8 quantization on newer NVIDIA GPUs. This is a known compatibility issue with CTranslate2 on newer architectures (sm_120 / Blackwell / RTX 50-series) where INT8 tensor cores require specific padding. The fix was recently merged into CTranslate2 master via PR #1982 (Enable multiple of 16 padding for INT8 Tensor Cores): Issue: https://github.com/OpenNMT/CTranslate2/issues/1865 Fix: https://github.com/OpenNMT/CTranslate2/pull/1982 Since you authored the fix in CTranslate2, could you please trigger a new build/release for whisper-standalone-win that includes these latest CTranslate2 changes? This would resolve the crash for users with RTX 50-series cards.