llama-bench: support --version to print build info

[asbelin]

Adds a --version\ command-line flag to llama-bench for parity with other llama.cpp tools (e.g. llama-perplexity) that support it.

Changes

• \llama-bench --version\ now prints build/version info via \llama_print_build_info(llama_version())\ (build number, commit, compiler, target) to stderr and exits 0
• identical behavior to the common --version\ option registered in \common/arg.cpp\ (used by llama-perplexity)
• documents the flag in \print_usage\

No new dependencies: \build-info.h, \llama.h, and <cstdlib>\ are already included by llama-bench.cpp. Base is v0.4.1; llama-bench.cpp is unchanged between v0.4.1 and master.

• I have read and agree with the contributing guidelines (https://github.com/ggml-org/llama.cpp/blob/master/CONTRIBUTING.md)
• AI usage disclosure: YES - I used an AI coding assistant to draft and review this small patch.

[nikwen]

Tested locally and it works:

nikwen@MacBook-Pro llama.cpp % ./build/bin/llama-bench --version      
version: 0.4.1-dev (build 10991, commit 930e2fa59)
built with AppleClang 21.0.0.21000334 for Darwin arm64

The output is identical to what llama-cli prints.