Preserve typed Playground command output for plugin-state verification

[chubes4]

Problem

BET-19 recipe setup succeeds in a prepared WPCOM/Playground runtime, including plugin activation, but final active_plugins verification fails after the result reaches the cached CLI boundary:

Unexpected token '\u0000' ... is not valid JSON

The wordpress.plugin-state command returns the raw PlaygroundRunResponse from server.playground.run(). That response can expose text through a lazily backed transport result. A later cache/CLI JSON parse can then observe NUL/binary bytes rather than the command's JSON payload.

#2048 / 789b5aa5 repaired the equivalent temporary WP-CLI wrapper path by snapshotting response.text before it crosses the cleanup boundary. Plugin-state does not take that path, so it remains vulnerable. Runtime service output aliases (#2147, #2228) are unrelated typed environment-binding repairs.

Expected contract

Materialize Playground command responses at the runtime boundary: preserve text, exitCode, and errors as plain typed values before they escape runPlaygroundCommand. Do not strip arbitrary bytes or retry JSON parsing. The plugin-state JSON producer continues to serialize the normalized active_plugins array.

Regression coverage

Add a focused runtime fixture whose direct Playground command response changes to NUL-containing output after the command boundary. Execute wordpress.ensure-plugin-active or wordpress.plugin-state, then assert the returned result is the original JSON and JSON.parse succeeds after the deferred invalid output becomes observable.

Scope checked

• GitHub open/closed issues and PRs: no BET-19, active_plugins, NUL, or output-alias repair covers this path.
• Local branches: fix/2048-wp-cli-json-result (789b5aa5) is merged into main and covers only temporary WP-CLI wrappers.
• verify/bet19-wpcom-runtime is the canonical SSI migration branch, not this bug.

AI assistance

• Model: OpenAI GPT-5.6 Terra
• Tool: OpenCode
• Used for: Investigated the runtime boundary and drafted this issue. Chris Huber remains responsible for review and prioritization.