Update PKGBUILDs

• Updated release PKGBUILD for v4.8.1
• Updated git PKGBUILD to simplify updating AUR (make both the same, avoiding errors when updating)
• Updated CI build script to disable CUDA build manually due to the change above

Once this is merged I will update AUR.

v4.8.1 builds fine but I cannot use STT (WhisperCPP either Small or Large-V3-Turbo, Vosk works fine, I did not try the git version) using CUDA, it gets stuck at "Transcribing audio file... 93%" :

[D] 18:29:39.677652645.677 0x7fa5f4dfe680 process_buff:588 - speech frame: samples=674400
[D] 18:29:39.677654660.677 0x7fa5f4dfe680 decode_speech:684 - speech decoding started
[D] 18:29:39.677657381.677 0x7fa5f4dfe680 decode_speech:700 - audio_ctx: 1500
[D] 18:29:39.687378377.687 0x7fa73060bd00 () - service refresh status, new state: transcribing-file
[D] 18:29:39.687403687.687 0x7fa73060bd00 () - task state changed: 1 => 2
[D] 18:29:39.687438220.687 0x7fa73060bd00 () - service refresh status, new state: transcribing-file
[D] 18:29:39.703829739.703 0x7fa73060bd00 () - app task state: speech-detected => processing
ggml_cuda_compute_forward: IM2COL failed
CUDA error: device kernel image is invalid
  current device: 0, in function ggml_cuda_compute_forward at /home/user/Projects/aur/dsnote.local/src/dsnote-4.8.1/external/whispercppcublas/ggml/src/ggml-cuda.cu:2341
  err
/home/user/Projects/aur/dsnote.local/src/dsnote-4.8.1/external/whispercppcublas/ggml/src/ggml-cuda.cu:70: CUDA error

Nvidia 3050 6GB

$ nvidia-smi --version
NVIDIA-SMI version  : 575.64
NVML version        : 575.64
DRIVER version      : 575.64
CUDA Version        : 12.9