[loki-distributed] loki-querier architecture does not allow KEDA ScaledObject

[siryur]

Chart version: loki-distributed v0.77.0

As the official grafana documentation suggests:

We recommend using Kubernetes Event-Driven Autoscaling (KEDA) to configure autoscaling based on Prometheus metrics. Refer to Deploying KEDA to learn more about setting up KEDA in your Kubernetes cluster.

The problem here is that whenever I want to implement a ScaledObject on my side,
I don't have a way to not set replicas nor deploy the default Kubernetes HPA.

See:

helm-charts/charts/loki-distributed/templates/querier/deployment-querier.yaml

Lines 15 to 17 in 9580ada

	{{- if not .Values.querier.autoscaling.enabled }}
	replicas: {{ .Values.querier.replicas }}
	{{- end }}

and

helm-charts/charts/loki-distributed/templates/querier/hpa.yaml

Line 1 in 9580ada

{{- if and .Values.indexGateway.enabled .Values.querier.autoscaling.enabled }}

---

So, in a case where I use this chart as a dependency of my (self-made) chart with:

dependencies:
- name: loki-distributed
  repository: https://grafana.github.io/helm-charts
  version: 0.77.0

And then I build a KEDA ScaledObject, I'm stuck because the chart will always deploy
a Deployment with spec.replicas or the kind: HorizontalPodAutoscaler object:

apiVersion: keda.sh/v1alpha1
kind: ScaledObject
metadata:
  name: {{ include "loki.querierFullname" $lokiContext }}
  namespace: {{ .Release.Namespace }}
  labels:
    {{- include "loki.querierLabels" $lokiContext | nindent 4 }}
spec:
# (...)

---

I suggest one of the following:

• We introduce the KEDA ScaledObject implementation on the chart
• We introduce an option to not set neither of those objects

[siryur]

Duplicated: #2126