[build] Add pre-push formatter hook

[AutomatedTester]

Summary

• Adds .githooks/pre-push — a repository-managed hook that runs ./scripts/format.sh --pre-push before every push.
• If formatters modify files the hook auto-commits them (chore: apply formatters before push) and aborts the push, asking the contributor to re-push with the clean state included.
• If the working tree is dirty the hook aborts immediately to avoid mixing uncommitted work with the formatter output.
• Documents the one-time setup step (git config core.hooksPath .githooks) and expected push behavior in CONTRIBUTING.md under Step 6: Push.

Motivation

CI runs ./go format and fails if git diff is non-empty afterwards. Contributors frequently hit this only after opening a PR. This hook catches the issue locally, creates the fix commit automatically, and surfaces a clear re-push message — removing the round-trip.

Reviewer notes

• The hook is opt-in: contributors must run git config core.hooksPath .githooks once per clone. Nothing changes for anyone who does not set this.
• SKIP_FORMAT_HOOK=1 guards the auto-commit path to prevent hook recursion.
• --no-verify remains available as an escape hatch (documented as discouraged since CI will still fail).
• The hook delegates all formatter logic to the existing scripts/format.sh; no formatter code is duplicated.

Test plan

• Clone the repo, set core.hooksPath, make a change that requires formatting, and verify the hook commits the fix and asks to re-push.
• Verify second push succeeds with no additional formatter commits.
• Verify a dirty working tree is rejected before any formatters run.
• Verify git push --no-verify skips the hook entirely.

[diemol]

[titusfortner]

I think I'm ok with a pre-push hook in theory, but...

1. we would need to support both sh and ps1
2. I don't like the auto-commit. The reason we have a auto-commit step in the CI is because it would take forever to get the failure have the submitter see the problem, fix the problem, rerun it, when it is a simple formatting thing. If we can alert the user real time to an issue, they can commit it themselves and submit, not a heavy lift.
3. I'm not sure we should assume that everyone looking to contribute has a working bazel setup

My proposal is that we just encourage people to create their own pre-push or pre-commit hooks.

My pre-push hook is just:

#!/usr/bin/env bash
./scripts/format.sh --pre-push --lint

A separate note, I kind of think --pre-push should default to --lint behavior