Update: card metrics query

[ibraheem-abe]

Summary

This PR updates the evaluation detail metrics to only take into account non disqualified agents.

It also updates the GET /evaluation-sets/{ID} endpoint to include the agent_id for the lowest average cost and lowest runtime:

...
  "efficiency": {
    "lowest_average_cost_usd_top_agents": {
      "agent_id": "c9d5efbc-99c3-58a0-9b00-764f0fc2c9f9",
      "value": 0.0368
    },
    "lowest_average_runtime_seconds_top_agents": {
      "agent_id": "b189c272-78ab-5de1-a169-03bb8788fa42",
      "value": 338.5569
    },
    "average_agent_cost_usd": 0.0947,
    "average_agent_runtime_seconds": 563.4667
  }
...

The GET /evaluation-sets/{ID}/leaderboard endpoint was updated so that cancelled agents do not have a rank and their final score reflects the projected maximum score.