build(make): harden kubectl wait to cover all applied CRDs in deploy targets

[coderabbitai]

Follow-up from PR #1150

Related comment: #1150 (comment)
Requested by: @bartoszmajsak

---

Three targets stand in the CI lane,
deploy, deploy-dev, deploy-ci by name.
Each waits on one CRD and moves along,
but the rest may lag — a race-prone song.

When multiple CRD sets are applied in full,
a single kubectl wait is not sufficient pull.
The controller may roll out before all CRDs are set,
and flaky failures are a debt we've not paid yet.

So harden the wait, let all CRDs be Established first,
quench the race condition's thirst.
Replace the lone wait with a list complete,
or use --all to make the cadence neat.

---

Problem

In all three targets (deploy, deploy-dev, deploy-ci), the kubectl wait --for=condition=established call gates on a single CRD even though multiple CRD kustomize overlays are applied beforehand. This creates a potential race condition in CI where other CRDs may not be Established before the controller is rolled out.

Suggested fix

Replace the single-CRD wait with either:

kubectl wait --for=condition=established --timeout=120s     crd/inferenceservices.serving.kserve.io     crd/trainedmodels.serving.kserve.io     crd/clusterservingruntimes.serving.kserve.io     crd/servingruntimes.serving.kserve.io     crd/inferencegraphs.serving.kserve.io     crd/clusterstoragecontainers.serving.kserve.io     crd/llminferenceservices.serving.kserve.io     crd/llminferenceserviceconfigs.serving.kserve.io

or a broader:

kubectl wait --for=condition=established --timeout=120s crd --all

Scope

• Makefile: deploy target
• Makefile: deploy-dev target
• Makefile.overrides.mk: deploy-ci target