Consistent score bands and terminal model picks

What changed

• Competency scores on a candidate report now use the same four bands in every card: Weak, Promising, Solid, and Strong.
• On assessments with a Claude Code or Codex question, candidates only see models the terminal agents can run, so the model they pick is the one their terminal uses.
• Each company workspace keeps its own Assessment Copilot conversation, so switching workspaces and coming back returns you to where you left off.