Skip OCI registry tests when Docker is unavailable

[rootp1]

Summary

• skip Docker-dependent pkg/oci tests cleanly when the Docker daemon is unavailable instead of aborting the whole package in TestMain
• add focused coverage for the Docker-unavailable detection logic

Linked Issue

Fixes #7008

What Changed

• changed the OCI test harness to record registry setup failures and let Docker-dependent tests skip individually when Docker is unavailable
• updated the Docker-backed OCI tests to use a shared requireOCIRegistry helper
• added unit coverage for Docker-unavailable error classification

Verification

• go test ./pkg/oci — passed
• go test -v ./pkg/oci — passed
• go test ./pkg/... ./cmd/... — failed: unrelated envtest prerequisite missing (/usr/local/kubebuilder/bin/etcd not found in pkg/app/piped/executor/kubernetes)
• make check — failed: yarn is not installed, so the required web build step could not run
• make lint/go MODULES=. — failed: docker is not installed, so the lint container could not start
• make check/dco — failed: the script checks commits since origin/master and stops on upstream commit 25ae69937e55e8330bff96c34346718f3e5e7148, which already lacks a sign-off in this local range
• git diff --check — passed

Scope

• only pkg/oci test files were changed; unrelated tracked files were not modified

What this PR does:

Skips Docker-dependent OCI registry tests cleanly when Docker is unavailable and keeps genuine setup failures visible.

Why we need it:

Plain Go test runs should not abort on machines that do not have a Docker daemon available.

Which issue(s) this PR fixes:

Fixes #7008

Does this PR introduce a user-facing change?:

• How are users affected by this change: contributor and CI test runs no longer hard-fail in pkg/oci solely because Docker is unavailable
• Is this breaking change: no
• How to migrate (if breaking change): not applicable