[SUBMISSION] wartzar-bee CI Cost Guardrail

### Title

wartzar-bee CI Cost Guardrail

### Description

A GitHub Action that estimates a PR's token-cost delta for AI-agent/prompt code (HEAD vs base), comments on the responsible files, and can block the build when cost regresses past a configurable threshold (default 20%). Deterministic — powered by the `@wartzar-bee/tokenscope` CLI; no LLM call, no data leaves CI. Catches prompt/context bloat before it ships.

### URL

https://github.com/wartzar-bee/ci-guardrail

### Category

Continuous Code Optimization

### Guidelines

- [X] Related to AI-powered automation for software collaboration — it guards the cost of AI-agent code in the PR loop.
- [X] Publicly accessible — repo and Marketplace listing both live, MIT.
- [X] Not already listed.

### Note on scope

Being upfront: this Action does not itself invoke an LLM — it counts tokens deterministically for AI-generated/prompt code. I read that as in-scope for Continuous Code Optimization alongside the existing entries (and the list already carries non-AI helpers under Utilities), but it's your call, and a "no" is a perfectly fine answer.

Start by reviewing the submitted repository and Marketplace listing at the provided URL, then compare the Action with this list’s existing entries and inclusion guidelines. Done means the submission is either added to the appropriate list section with accurate details or its acceptance decision is recorded.

Written by the indexing model from the issue text.