llama.cpp PR adds support for DeepSeek V4

A pull request by am17an adds initial support for running DeepSeek V4 locally via llama.cpp. The community effort enables inference for the model on consumer hardware.