End-to-end tests on a local model, and in CI on every PR

- Every e2e role runs on qwen3.8-27b on the RTX 4090 (e2e_roles();
  NISSE_E2E_MODEL overrides, e.g. claude-haiku-4-5). The first full run
  passed 7 of 8 in 26 minutes for $0.0000; the eighth was a real finding,
  fixed with the new-project PR.
- scripts/e2e.sh reports a run's spend on the test key and the day's total
  against its budget: above $0 on the local model means a paid fallback.
- CI runs the suite as its own job on pull requests (not pushes to main),
  on a throwaway sshd (--local), cancelling a run a newer push supersedes,
  so the checks job still reports in minutes.
- The plan says so: e2e on every PR; locally only -k / --lf.

Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01PPQ14L3LikyefCMFBqEbKE

tests/integration/test_end_to_end_loop.py

@@ -21,7 +21,7 @@ from pathlib import Path
 from typing import Any
 
 import pytest
-from e2e_support import dependent_task, fresh_repo, operator_gitea, task
+from e2e_support import dependent_task, e2e_roles, fresh_repo, operator_gitea, task
 
 from nisse_dev.config import AgentConfig, LoopTuning
 from nisse_dev.execution import SSHHost
@@ -74,6 +74,7 @@ def running(
             build_host=build_host(ssh_host),
         ),
         loop="new",
+        roles=e2e_roles(),
         state_db=keep / "state.db",
         # Short, so a bug's grace period runs out within the test.
         tuning=replace(live_config.tuning, loop=LoopTuning(bug_grace_s=BUG_GRACE_S)),