feat: track and report per-round retrieval metrics (attempts, failures, coverage)

[Hany-Almnaem]

Summary

This PR introduces internal tracking of per-round retrieval metrics in the SPARK module to support better visibility into network coverage and reliability.

##Changes include:

• Added a new lib/metrics.js module to track:
  • Total retrieval attempts per round
  • Retrieval failures per round
  • Unique (PayloadCID, SP) pairs tested per round
• Integrated retrievalMetrics into spark.js:
  • Calls recordRetrieval() and recordFailure() within nextRetrieval()
  • Resets and reports metrics on each task loop in run()
• Added a unit test: test/retrieval-metrics.test.js
  • Verifies counts, deduplication, and round indexing behavior

These metrics give insight into:

• How much of the Filecoin network is being sampled by SPARK
• How effective those retrieval attempts are
• Which trends might indicate network degradation

This groundwork will later support building Grafana dashboards

Related Issue

Resolves #34

[bajtos]

As mentioned on Slack, we will be sunsetting the permissionless Spark Checker node soon.