IN-6457: add DI metrics aggregation and whitelist for Insights

[sk09-dev]

This change adds support in do-agent to properly collect and aggregate Dedicated Inference (DI) metrics for Insights.

this PR:

• Adds DI metrics to the whitelist so they are accepted by do-agent
• Defines aggregation rules for DI metrics to control label cardinality
• Preserves Prometheus histogram semantics by keeping le labels for _bucket metrics
• Drops high-cardinality labels like job, pod, and OTEL scope labels where appropriate
• Ensures _bucket, _sum, and _count metrics remain consistent and query-safe
• The goal is to make DI metrics safe, efficient, and compatible with Insights ingestion.

Testing performed

Ran do-agent with a local DI Prometheus-style metrics endpoint:

go run ./cmd/do-agent
--di-metrics-address http://127.0.0.1:9109/metrics
--stdout-only

• Verified that DI metrics are:
• Successfully scraped
• Present in agent output
• Correctly aggregated
• Confirmed that:
• Histogram metrics (*_bucket, *_sum, *_count) are emitted correctly
• le label is preserved for bucket metrics
• High-cardinality labels (job, pod) are removed from aggregated DI metrics
• Manually inspected output to ensure no unintended label or metric drops
• No integration or production environment changes were required for this PR.