fix(monitor): keep regions visible before scheduled runs

[pavelm]

Summary

Freshly deployed regional workers now remain visible to Prometheus and Grafana while waiting for their staggered benchmark slot. Startup publishes zero-valued attempt and failure counter series for every configured provider and network, preserving both the nonce-safe schedule and real counter semantics. This prevents region counts and dashboard variables from disappearing for up to roughly 70 minutes after a worker restart, while also giving the first completed batch a valid counter baseline.

Validation

• bun test — 329 passed, 0 failed, 768 assertions
• bunx tsc --noEmit
• git diff --check

Production scrape labels are attached externally and were not available locally. After deployment, count(count by (region) (txe_bench_attempts_total{measurement_epoch="alchemy-status-v2",environment="prod",network="base-mainnet"})) should report all four workers without waiting for their benchmark batches to finish.

---