Incorrect high_water_mark / low_water_mark values for non-empty topic from KMinion

[derkhin]

🐞 Issue Title:

Incorrect message count reported for non-empty topic in Redpanda KMinion

---

🐳 Environment

• Image: docker.io/redpandadata/kminion:latest
• Image ID: docker.io/redpandadata/kminion@sha256:6724ee42c6f5fbf4f3bc90ce60f7172eac112c8063d4848d61201f69e7457e1e

---

📝 Description

For the topic fynd-json-shipment-bag-status, KMinion is reporting:

As a result, the derived message count (high - low) is 0 for all partitions.

---

🆚 Expected vs. Actual Behavior

• Expected:
  message_count = high_water_mark - low_water_mark
  This should reflect the actual number of retained messages per partition, consistent with the Kafka UI.

• Actual:
  Even though cleanup_policy="delete" and new messages are continuously published to the topic, the reported message count remains 0 in all partitions based on Prometheus/Thanos metrics.

---

📸 Visual Evidence

• Thanos Query Output - ( how high_water_mark & low_water_mark values are same ? )

---

✅ Cleanup Policy

• cleanup_policy="delete"

---

🔄 Notes

• The discrepancy persists across multiple scrapes and after publishing fresh messages.
• Prometheus/Thanos setup and label filters have been verified.

---

Please let me know if further logs or configuration details are needed.

[ElfoLiNk]

#298