fix(research): keep retained signal validation diagnostic

[Pigbibi]

Summary

• Treat validation of the retained signal as a diagnostic, not a prerequisite for generating independent new research context.
• Emit a fixed warning on failure without exposing validator output.
• Preserve existing artifacts and all new-artifact validation, provenance, time, and execution restrictions.

Verification

• Tests-first: original workflow failed both legacy-v1 and invalid-v2 cases (2 failed / 1 passed).
• Real validator and context CLI exercised using local synthetic prices only; retained bytes unchanged.
• Focused tests: 27 passed. Full suite: 97 passed. Ruff and git diff checks passed.
• Push/PR hooks do not trigger shadow research, model calls, or provider acquisition. No real dispatch, deployment, or model evaluation was performed.

Scope

Only dispatch_shadow_signal.yml and its existing workflow test file. No new dependencies or validators.