llama.cpp install paths split Mac users

Hot take: for most Mac users, Homebrew is the best default, and compiling only becomes worthwhile when you need bleeding-edge changes or very specific build options.

• –The official project docs list four valid paths: Homebrew, Docker, prebuilt releases, and source builds, so there is no single "correct" install method.
• –On macOS, Metal is enabled by default in source builds, so compiling is mostly about control and freshness, not some hidden performance unlock.
• –Prebuilt binaries are the fastest way to try llama.cpp because they skip toolchain setup and let you launch immediately.
• –Homebrew is the least annoying for ongoing use because upgrades are handled by the package manager; you normally do not have to manually delete old installs.
• –If you want the newest upstream behavior the same day it lands, source builds or release binaries will usually beat packaged formula updates.