Provide older-glibc Linux CUDA CLI build and optional Vulkan CLI fallback

[CKwasd]

Hi, thanks for CrispASR.

I tried running the latest Linux CUDA CLI release in Google Colab:

crispasr-linux-x86_64-cuda.tar.gz

The binary fails to start with:

/lib/x86_64-linux-gnu/libstdc++.so.6: version `GLIBCXX_3.4.32' not found
/lib/x86_64-linux-gnu/libstdc++.so.6: version `GLIBCXX_3.4.31' not found
/lib/x86_64-linux-gnu/libc.so.6: version `GLIBC_2.38' not found

Building CrispASR from source inside Colab works as a workaround, but it adds startup time and complexity for hosted GPU workflows.

Would you consider one or both of the following release options?

1. A Linux x86_64 CUDA CLI build produced on an older baseline such as Ubuntu 22.04, to avoid requiring GLIBC_2.38.
2. A Linux x86_64 Vulkan CLI build, useful as a fallback in environments where CUDA release binaries are incompatible.

For Google Colab specifically, option 1 is probably more important because Colab provides NVIDIA GPUs, while Vulkan support is not guaranteed in the headless runtime.

It would also help if the release notes documented the minimum required glibc and libstdc++ versions for Linux binaries.

Thanks.

[CrispStrobe]

All Linux release jobs now target ubuntu-22.04 (glibc 2.35) instead of 24.04 — should resolve the Colab GLIBC_2.38 error. A new crispasr-linux-x86_64-vulkan.tar.gz build was also added for non-CUDA GPU environments. Changes land in the next release tag.