Fix release scripts to support patch release workflow

[eguzki]

Problem

The current release scripts in the main branch do not properly support the patch release workflow, causing test failures when preparing patch releases on release branches. This issue affects both limitador-operator and dns-operator, both of which use similar release script patterns.

When attempting to prepare a patch release on a release branch:

• make verify-bundle fails due to hardcoded variables in make/release.mk
• Image building from release branches fails because variables cannot be overridden
• Tests expect the ability to customize VERSION and IMAGE_TAG for patch releases

This issue was discovered when attempting to prepare patch release v0.17.2 (see #253).

• Document the patch release workflow in release documentation

Current State Across Kuadrant Repos

limitador-operator:

• Main branch: Uses = assignment in generated make/release.mk ❌
• release-0.17 branch: Fixed in commit abefe4a to use ?= assignments ✅

dns-operator:

• Main branch: Uses = assignment in prepare-release target (same issue) ❌
• Release branches: Uses = in make/release.mk (see release-0.3/make/release.mk) ❌

authorino-operator:

• Uses a different approach with build.yaml instead of release.mk
• Variables are handled differently through yq and separate targets

Acceptance Criteria

Core Requirements

• make verify-bundle passes on release branches
• Image building works correctly from release branches

Alignment

• Consider applying the same fix to dns-operator (or document why approaches differ)

References

• PR Prepare release v0.17.2 #253: Initial patch release attempt that revealed the issue
• Commit abefe4a: The fix applied to release-0.17 branch
• dns-operator issue: Has the same problem in main branch prepare-release target
• Related: Kuadrant components should follow consistent patch release workflows