debug(vtex): probe isolate reuse + per-request timing

[viktormarinho]

Summary

Diagnostic — not for keeps. Wrap `runtime.fetch` in vtex's main.ts with a request counter and isolate-age log so we can see in `wrangler tail` whether Workers is reusing isolates across requests at GRU.

The runtime cache fix (decocms/studio#3300, vtex#427) is deployed but `tools/list` is still 4-7s on every call. Either:

• The cache wrapper isn't installing (we'd see no `[runtime]` logs from the wrapper)
• The cache installs but the closure resets per request (Workers spawning fresh isolates → `req#1` repeats)
• Something else dominates per-request CPU

This probe answers question 2.

Test plan

• Merge → deploy
• Hit `tools/list` 5x in a row, check `wrangler tail` for `[vtex-probe]` lines
• If counter resets to 1 every call → isolate not reused → revisit caching strategy
• If counter increments → cache should be hitting; runtime needs further investigation

🤖 Generated with Claude Code

---

Summary by cubic

Adds a temporary debug probe that wraps runtime.fetch in vtex/server/main.ts to log a per-request counter, isolate age, and request timing. This helps confirm Workers isolate reuse and diagnose why tools/list stays slow despite the cache fix.

^{Written for commit 804a736. Summary will update on new commits.}

[viktormarinho]

abandoning workers migration, reverting to kubernetes-bun