tf-plan/tf-apply: allow extra Terraform CLI arguments such as -refresh=false #170

What

tf-plan.yaml and tf-apply.yaml have no input for extra Terraform CLI arguments, so a caller cannot pass -refresh=false (or -target, -parallelism, ...) to the plan and apply.

Where

• .github/workflows/tf-plan.yaml (tf-plan-v0): the plan step uses dflook/terraform-plan; the only hook is tf_pre_run, evaluated inside the action before Terraform runs.
• .github/workflows/tf-apply.yaml (tf-apply-v0): the same for dflook/terraform-apply.

Why it matters

In DND-IT/tamedia-ai-platform, a version promotion only changes one pin in a stack's environments/<env>.tfvars, but every apply refreshes the whole stack (stacks/agentcore, about 150 resources and modules), which is most of a 3 to 14 minute run. Skipping the refresh for pin-only changes would cut that substantially. See DND-IT/tamedia-ai-platform#541 for the rest of the promotion speed-up.

Suggested fix

Add an optional input, e.g. tf_cli_args, to both workflows and pass it through to the plan and apply, either as an action input if dflook/terraform-plan / terraform-apply support one, or as TF_CLI_ARGS_plan / TF_CLI_ARGS_apply in the step environment. Check that dflook/terraform-apply's comparison against the pull request's plan still matches when both sides skip the refresh.