Bugfix: Allow scale-down when at one-shard-per-node

[otrosien]

One-line summary

Issue: Due to a wrong edge-case handling, when hitting a one-shard-per-node setup, the es-operator will not scale-down further.

Closes #398

Description

There have been reports of es-operator being stuck during scale-down. It happens in this situation:

• At maximum scaling we have one-shard-per-node
• There is a scaling hint "DOWN" that doesn't execute, as the edge-case of having less than one shard per node (which was meant to avoid over-scaling at scale-up) is triggered.

Changes

• Instead of stopping the scaling, ensure that the scaling respects a minimum value of one-shard-per-node.
• Fixes the e2e test which were failing due to duplicate env variable

Types of Changes

• Bug fix (non-breaking change which fixes an issue)

[zalando-compr-opensource]

invalid team ID

The team ID in your .zappr.yaml file does not appear to be valid. Please, fix
this before team ID checks will be added back into ComPR's specification check.

You can follow this guideline for help.

[otrosien]

👍

[hooseins]

👍