feat: add Maiat trust evaluator example for ACP jobs

[JhiNResH]

Summary

Adds an example showing how to use Maiat as a trust-based evaluator for ACP jobs via the on_evaluate callback.

What it does

The example provides two integration options:

Option 1: One-liner (recommended)

from maiat_evaluator import maiat_evaluator

acp_client = VirtualsACP(
    ...,
    on_evaluate=maiat_evaluator(),  # that's it
)

Option 2: Inline (no extra dependency)

Copy the inline evaluator function — zero dependencies beyond httpx.

Evaluation Logic

1. Garbage detection — Rejects empty, too-short, or placeholder deliverables (hello, {}, etc.)
2. Trust score check — Queries Maiat's trust oracle (18,000+ agents indexed)
3. Auto-approve/reject — Score >= 80 auto-approves, < 30 rejects
4. Outcome recording — Reports results back for trust score updates

Why

Currently there's no example of a trust-based evaluator in the GAME SDK examples. Most agents use a simple job.evaluate(True) which auto-approves everything — including garbage deliverables. This example shows how to add meaningful quality checks.

Files Changed

• examples/game/maiat_evaluator_example.py (new)