feat(cli): --version reports the commit summit was built from

[samlaf]

Closes SEI-845.

Need summit's version when founding a TEE network since we pin the genesis format, which can change over different summit versions. Seismic-images release pin a summit version that is used in the image, so goal is to eventually have the seismic-tee CLI download the same summit version from summit releases.

Having summit document it's version is needed to help debug version mismatches.

People recommend using vergen (just like reth does) but decided against it because it's very easy to roll our own with LLMs nowadays, so we keep our supply chain smaller. Plus vergen seems to have issues when building dirty worktrees (see bullet point below)

LLM Summary

summit --version follows rustc and cargo (not seismic-reth, whose -V and --version print different things): one line by default, key: value lines with -vV for scripts:

summit 0.0.3-alpha (7aac73b40 2026-10-10)
release: 0.0.3-alpha
commit-hash: 7aac73b40db1a3693da3c1970d32485381d04cf6
commit-date: 2026-10-10
host: x86_64-unknown-linux-gnu

• Dirty: the hash gets -dirty when tracked files differ from the commit, so a dirty build never matches a pin.
• Reproducible: every value comes from the inputs (commit, its commit date, target), none from the build itself (no timestamp or path).
• Needs git: without git or a checkout of this workspace, the hash and date read unknown, never a wrong commit. Release CI and the image build from git clones.
• Not vergen: it reruns only when HEAD or the branch ref moves, so its dirty flag goes stale after an edit, and its reproducibility settings target build timestamps we don't embed. node/build.rs has no dependencies and reruns on any tracked-file edit.
• clap's version flag can't take -v, so -V/-v are declared on CliArgs; summit prints the version when no subcommand is given.

The release workflow now fails unless commit-hash: equals the commit it built.

[samlaf]

Edited the docs describing the behavior of --dirty to not include untracked files after chatting with @matthias-wright , reasoning being: