[DOCS] Llama-CPP Linux NVIDIA GPU support and Windows-WSL

[hamzahassan66]

Description

What:

Replace
CMAKE_ARGS='-DLLAMA_CUBLAS=on' poetry run pip install --force-reinstall --no-cache-dir llama-cpp-python numpy==1.26.0
with
CMAKE_ARGS='-DGGML_CUDA=on' poetry run pip install --force-reinstall --no-cache-dir
llama-cpp-python==0.2.90 numpy==1.26.4 markupsafe==2.1.5

Why:

-DLLAMA_CUBLAS is deprecated
-dependency versions for compatibility

[giswa]

Please add this to the docs.