Bug: 0.10.0 release breaks GPU on older OSs

### Contact Details _No response_ ### What happened? Previously, it worked with GPU acceleration: ``` user@pop-os:~$ ~/.local/opt/llamafile/llamafile-0.9.3 -m ~/llms/tinyllama-1.1b-chat-v1.0.Q2_K.gguf --server --host '::' --port 8081 --log-disable -ngl 999 {"function":"server_params_parse","level":"INFO","line":2737,"msg":"logging to file is disabled.","tid":"12988640","timestamp":1775857540} import_cuda_impl: initializing gpu module... get_rocm_bin_path: note: hipcc not found on $PATH get_rocm_bin_path: note: $HIP_PATH/bin/hipcc does not exist get_rocm_bin_path: note: /opt/rocm/bin/hipcc does not exist link_cuda_dso: note: dynamically linking /home/user/.llamafile/v/0.9.3/ggml-rocm.so link_cuda_dso: warning: libamdhip64.so.6: cannot open shared object file: No such file or directory: failed to load library link_cuda_dso: note: dynamically linking /home/user/.llamafile/v/0.9.3/ggml-rocm.so link_cuda_dso: warning: libamdhip64.so.6: cannot open shared object file: No such file or directory: failed to load library get_nvcc_path: note: nvcc not found on $PATH get_nvcc_path: note: $CUDA_PATH/bin/nvcc does not exist get_nvcc_path: note: /opt/cuda/bin/nvcc does not exist get_nvcc_path: note: /usr/local/cuda/bin/nvcc does not exist link_cuda_dso: note: dynamically linking /home/user/.llamafile/v/0.9.3/ggml-cuda.so ggml_cuda_link: CUDA kernel version 12.7 ggml_cuda_link: CUDA runtime version is 12.8 (!!!) ggml_cuda_link: welcome to CUDA SDK with tinyBLAS ---snip--- ``` The new version doesn't: ``` user@pop-os:~$ ~/.local/opt/llamafile/llamafile-0.10.0 -m ~/llms/tinyllama-1.1b-chat-v1.0.Q2_K.gguf --server --host '::' --port 8081 --log-disable -ngl 999 cuda: /lib/x86_64-linux-gnu/libstdc++.so.6: version `GLIBCXX_3.4.32' not found (required by /home/user/.llamafile/v/0.10.0/ggml-cuda: failed to load library cuda: /lib/x86_64-linux-gnu/libstdc++.so.6: version `GLIBCXX_3.4.32' not found (required by /home/user/.llamafile/v/0.10.0/ggml-cuda: failed to load library warning: no usable GPU found, --gpu-layers option will be ignored warning: one possible reason is that llama.cpp was compiled without GPU support warning: consult docs/build.md for compilation instructions ``` (No further output; at this point I can interact with it, but nvtop shows no GPU use and tokens/second shows it much slower in a way that I would expect on CPU) And indeed my system doesn't have GLIBCXX_3.4.32: ``` user@pop-os:~$ grep PRETTY /etc/os-release PRETTY_NAME="Pop!_OS 22.04 LTS" user@pop-os:~$ ls /lib/x86_64-linux-gnu/libstdc++* /lib/x86_64-linux-gnu/libstdc++.so.6 /lib/x86_64-linux-gnu/libstdc++.so.6.0.30 ``` (This version is a bit old, but it's still supported, and the next release was only quite recently released: https://endoflife.date/pop-os ) ### Version llamafile v0.10.0 ### What operating system are you seeing the problem on? Linux ### Relevant log output ```shell ```