Specshift badges

Every run on /eval saves its scorecard under a report id, shown on the report page. Add the overall score from that run to a README with this snippet, replacing <report-id> with your id:

[![Specshift](https://pragma.wentzel.ai/api/badge/<report-id>.svg)](https://pragma.wentzel.ai/eval/<report-id>)

The badge shows the score saved with that report and its methodology version. It does not update on its own: re-run the eval and use the new report id to refresh it. Unknown report ids return 404.