CUDA addon

[MaurUppi]

I noticed that SmartSub has a lot of CUDA versions, and the addon is from a forked Whispers.cpp repo.

Questions:

• What reasons drive you to maintain three branches?
• optimized node pretty huge (748 MB), I don't think you packed in the App.
• The latest release has addon-macos-arm64-coreml.node, and is identical (same SHA) to addon-macos-arm64.node, any reasons?

addon-windows-cuda-1180-generic.node
addon-windows-cuda-1180-optimized.node
addon-windows-cuda-1220-generic.node
addon-windows-cuda-1220-optimized.node
addon-windows-cuda-1241-generic.node
addon-windows-cuda-1241-optimized.node

SmartSub_Mac_2.5.2_x64.dmg

• sha256sum addon.coreml.node
  28267749187f3a57bdab426cafc2080506fa5931feec5477084295827ccf378b
• sha256sum addon.node
  f3d02118321225cf9762e1735280936d7a5c2cc5ca5b6d04d5632a97012bae54

whisper.cpp builds 20250702-094405
addon-macos-arm64-coreml.node sha256:28267749187f3a57bdab426cafc2080506fa5931feec5477084295827ccf378b
addon-macos-arm64.node sha256:28267749187f3a57bdab426cafc2080506fa5931feec5477084295827ccf378b
addon-macos-x64.node sha256:f3d02118321225cf9762e1735280936d7a5c2cc5ca5b6d04d5632a97012bae54

[buxuku]

• Because different CUDA versions have certain compatibility and higher versions have some new features, support for three common CUDA versions is provided
• Although the original addon.node file is large, it has a relatively high compression ratio. After packaging, it is compressed. You can extract the installation package file to verify it
• Thank you for the reminder. This is a bug. I built two different versions of the addon.node file on the same runner, but during the second build, it reused the build cache from the first one, thus resulting in the same build product. I will fix it as soon as possible.