Simplify benchmark deployments and improve result reporting

[gabotechs]

Summary

• remove the runtime-file handoff and capture published engine artifacts directly from standard command output
• make benchmark iteration and warmup settings explicit
• improve result comments with exact base/head commit links, a source comparison link, and expandable verification and timing details

Why

The runtime metadata file added unnecessary state and a second handoff mechanism to deployments. Engine publishing and deployment can instead behave like normal shell commands: publishers return their artifact identifiers directly to the caller.

Benchmark comments also need enough immutable source and execution detail for reviewers to verify what was actually tested.

Validation

• npm run format:check
• npm run build
• npm test (100 tests)
• bash -n for all changed shell scripts
• git diff --check
• live TPCH/SF1 benchmark-bot run against Add dynamic stage built events datafusion-contrib/datafusion-distributed#570, including build, deployment, comparison report, and Helm cleanup