chore(metrics): add metric about opendal retries count

[flaneur2020]

I hereby agree to the terms of the CLA available at: https://docs.databend.com/dev/policies/cla/

Summary

Retry count is a useful metric which may help us detect unhealthy conditions in network or storage. Sometimes it might not become an incident, but many retries may signal stability decline, the earlier we notice about this, the more chance for us to prevent incidents ealier.

Tests

• No Test - add a single metric

Type of change

• Other (please describe): metrics change

---

This change is