feat(prometheus): add histogram_Mode toggle for classic, native, and dual ingestion.

[jacobhallburns]

Description

This implementation introduces a histogramMode toggle within prometheus.libsonnet, providing the universal control requested in Issue #2784. By moving away from legacy bucket-based histograms at the library level, the solution maintains the integrity of the kube-prometheus manifests as the source of truth.

The implementation has been refactored to follow the maintainer's feedback: intermediate variables have been removed, and the logic is now applied directly within the resource specifications.

Key Features:

• Operational Modes: Supports classic (default), native (high-resolution sparse histograms), and dual (parallel ingestion).
• Library-Level Implementation: Uses std.map to ensure consistent histogram settings across all ServiceMonitor endpoints while maintaining a DRY codebase.
• Migration Safety: The dual mode allows for parallel ingestion of both formats, ensuring backward compatibility for existing dashboards during the transition.

Closes #2784

Type of change

• CHANGE (fix or feature that would cause existing functionality to not work as expected)
• FEATURE (non-breaking change which adds functionality)
• BUGFIX (non-breaking change which fixes an issue)
• ENHANCEMENT (non-breaking change which improves existing functionality)
• NONE (if none of the other choices apply. Example, tooling, build system, CI, docs, etc.)

Development & Testing Guide

To provide a reproducible testing environment, I utilized a Kubernetes-in-Docker (Kind) cluster and a specialized toolchain container.

Initial Setup

# Build the toolchain image (contains jsonnet, jb, and gojsontoyaml)
docker build -t kube-prom-builder .

# When building the initial Docker, or after each histogramMode change
docker run --rm -v $(pwd):/app --entrypoint "/bin/bash" kube-prom-builder -c \
"jb install && \
jsonnet -J /app/vendor -J /app/jsonnet -m manifests example.jsonnet | \
xargs -I{} sh -c 'cat {} | gojsontoyaml > {}.yaml'"

# Create the Kind cluster
kind create cluster --name prometheus-test

# Deploy base manifests (Apply CRDs first)
kubectl apply --server-side -f manifests/setup/
kubectl apply -f manifests/

# Monitor the rollout
kubectl rollout status statefulset prometheus-k8s -n monitoring

# Port Forwarding
kubectl port-forward -n monitoring svc/prometheus-k8s 9090:9090

Changelog entry

Add histogramMode toggle to support classic, native, and dual histogram ingestion modes across the stack.

[simonpasquier]

why do we need these?

[jacobhallburns]

Removed. These were redundant intermediate variables; the logic has been moved directly into the resource specifications to simplify the component's internal API.

[simonpasquier]

Can we avoid the intermediate variables and move this where it's used directly?

[jacobhallburns]

Done. I have eliminated the intermediate variables and moved the logic directly into the Prometheus spec and ServiceMonitor endpoints.

I used std.map for the ServiceMonitor endpoints to keep the implementation DRY and consistent. I've verified the fix in a local kind cluster—the generated configuration correctly enables scrape_native_histograms across all targets and I've confirmed native samples are being ingested in the TSDB.