Make ExecCommandOnPod() context-aware in e2e pods test helper

Summary

Make ExecCommandOnPod() in test/e2e/performanceprofile/functests/utils/pods/pods.go context-aware so callers can bound exec calls with a timeout/cancellation, instead of the helper internally using exec.StreamWithContext(context.TODO(), ...).

Rationale

During review of PR #1556, it was noted that new GOMAXPROCS-related e2e tests in test/e2e/performanceprofile/functests/1_performance/performance.go call pods.ExecCommandOnPod(), which is unbounded because the underlying helper uses context.TODO(). This means a stalled SPDY exec session could hang indefinitely, even though the pod cleanup path (DeferCleanup) was already fixed to use context.WithTimeout(...).

Since ExecCommandOnPod() is a shared utility used across many e2e tests, changing its signature/behavior was deferred to a follow-up to avoid unintended impact on other tests.

Affected areas

• test/e2e/performanceprofile/functests/utils/pods/pods.go (ExecCommandOnPod implementation)
• Call sites across test/e2e/performanceprofile/functests/... that use ExecCommandOnPod()

Acceptance criteria

• ExecCommandOnPod() (or a new context-aware variant) accepts a context.Context parameter and passes it through to exec.StreamWithContext(...) instead of using context.TODO() internally.
• Existing call sites are updated to pass an appropriate context (e.g., with a timeout), or a backward-compatible wrapper is provided to avoid breaking existing tests.
• No regression in existing e2e test suites that rely on this helper.

References

• PR: https://github.com/openshift/cluster-node-tuning-operator/pull/1556
• Discussion: https://github.com/openshift/cluster-node-tuning-operator/pull/1556#discussion_r3528568934

Requested by: @mrniranjan