What type of PR is this?

/kind bug

What this PR does / why we need it:

The PR changes Horizontal Pod Autoscaler behavior: TL;DR we're adding logic that if more than one HPAs are operating on the same pod, we stop them.

There are cases when selected Pod sets overlap due to misconfiguration:

Example 1:

Deployment A (with controlling HPA A) has selector:

...
spec:
  selector:
    matchLabels:
      pod: a
...

Deployment B (with controlling HPA B) has selector:

...
spec:
  selector:
    matchLabels:
      pod: b
...

And there is a group of Pods with:

...
labels:
  pod: a
  pod: b
...

Example 2:

HPA A and HPA B point to the same deployment.

As a result:

The same Pods are actuated first by HPA A then by HPA B in an unpredictable way.

Intended behavior:

In this case the HPAs should be disabled and the problem — reported to the user through conditions.

The algorithm to determine these situations is as follows:

1. For current HPA list the selected pods.
2. For each pod check whether it is selected by other HPAs too.
3. If so then set ScalingActive=false and return an error.

Which issue(s) this PR fixes:

Fixes #112027

Special notes for your reviewer:

The BiMultimap from package selectors is used to store all active HPAs' selectors and efficiently find the HPA-references to current pods.

The BiMultimap operates on:

• objects — or in our case Pods
• selectors — HPAs with selectors

It supports methods: Put (object), PutSelector — to insert entities; and Select — to get objects by a given selector; ReverseSelect — to get selectors that point to a given object.

Does this PR introduce a user-facing change?

Added selector validation to HorizontalPodAutoscaler: when multiple HPAs select the same set of Pods, scaling now will be disabled for those HPAs with the reason `AmbiguousSelector`. This change also covers a case when multiple HPAs point to the same deployment.

Additional documentation e.g., KEPs (Kubernetes Enhancement Proposals), usage docs, etc.: