feat(mcp): add benchmark_me and rank_if — field-relative tools

Two new remote HTTP MCP tools that answer the human questions:

- benchmark_me: "How good am I?" — benchmarks your cascade against
  the live leaderboard. Returns percentile, estimated rank, distance
  from median/top 10%/top 1%, strongest and weakest metric vs field
  median, and a one-line interpretation. Accepts pillars or a codename.

- rank_if: "What would it take to reach top 10%?" — counterfactual
  rank simulator. Takes your pillars + a target percentile, fetches
  the live board, computes the yield threshold, then simulates 5
  strategies (increase cache reads, decrease input, increase output,
  balanced, enable caching) and returns the smallest change needed
  to reach the target, with simulated rank/percentile/class.

Both tools fetch the live leaderboard for field statistics, making
them field-relative rather than pure-math. Still read-only, no auth.

Remote endpoint now exposes 10 tools (was 8). Updated mcp.json
discovery card, /mcp page metadata, FAQ JSON-LD, and tool listings.

Generated with [Devin](https://devin.ai)

Co-Authored-By: Devin <158243242+devin-ai-integration[bot]@users.noreply.github.com>

Author: SunrisesIllNeverSee

app/.well-known/mcp.json/route.ts

@@ -44,6 +44,8 @@ export async function GET() {
       "suggest_improvements",
       "self_improve",
       "rank_windows",
+      "benchmark_me",
+      "rank_if",
     ],
     install: "npx sigrank",
     docs: `${SITE_ORIGIN}/developers`,