Add runner-aware remote discovery to bench list

[chubes4]

Problem

When a Lab benchmark reports discovered ids: <none>, Homeboy has no supported command to run the same scenario-discovery contract on the selected runner.

The obvious diagnostic command is rejected:

homeboy bench list static-site-importer \
  --path /path/to/worktree \
  --placement lab \
  --runner homeboy-lab

error: unexpected argument --placement found

tip: to pass --placement as a value, use -- --placement

homeboy bench list --help exposes --path, --extension, and --rig, but no --runner or --placement. Passing these after -- would send them to the workload rather than inspect discovery remotely, so the generated tip is not a valid remediation.

This forced manual runner-job log inspection while diagnosing browser verification for Automattic/blocks-engine#659.

Expected

bench list should support the same execution environment selection as bench and report what the selected runner can actually discover after materialization/hydration.

Suggested shape:

homeboy bench list static-site-importer \
  --runner homeboy-lab \
  --path /path/to/worktree \
  --rig static-site-importer-fixture-matrix

The output should include controller and runner workload identities, extension/runtime versions, hydration state, discovered IDs, and bounded discovery diagnostics.

Acceptance criteria

• bench list accepts --runner and placement policy consistently with bench.
• Local paths and path settings are materialized/remapped exactly as execution would do.
• Discovery runs against the selected runner runtime and extension set.
• The command remains read-only and does not execute benchmark workloads.
• Unsupported-option guidance does not recommend forwarding Homeboy options through workload passthrough argv.

AI assistance

• AI assistance: Yes
• Tool: OpenCode with OpenAI GPT-5.6 Sol
• Used for: Reproduced the CLI gap and drafted the operator contract with Chris.