Enhance Selectel balance handling and error messaging in GPU rental system

- Updated Selectel documentation to clarify error responses related to balance issues.
- Introduced `peek_balance_rub` function for best-effort balance checks, handling API failures gracefully.
- Improved error messages for insufficient funds and quota issues, specifying Selectel's 403 policy response.
- Added balance checks in the doctor command to ensure users are informed about their balance status before attempting GPU creation.
- Refactored exception handling in cloud operations to provide clearer feedback on balance-related errors.

Co-authored-by: Cursor <cursoragent@cursor.com>

tests/test_up_stop_on_fail.py

@@ -3,7 +3,14 @@ from types import SimpleNamespace
 from gpu_rent.cli import _stop_after_failed_up
 
 
-def test_stop_after_failed_up_calls_cmd_stop(monkeypatch):
+def _quiet(monkeypatch):
+    monkeypatch.setattr("gpu_rent.cli.warn", lambda *_a, **_k: None)
+    monkeypatch.setattr("gpu_rent.cli.ok", lambda *_a, **_k: None)
+    monkeypatch.setattr("gpu_rent.cli.err", lambda *_a, **_k: None)
+    monkeypatch.setattr("gpu_rent.cli.log", lambda *_a, **_k: None)
+
+
+def test_stop_after_failed_up_skips_when_no_server(monkeypatch):
     calls: list[dict] = []
 
     def fake_stop(cfg, *, no_pull=False, log=None, destroy_disks=False):
@@ -11,13 +18,28 @@ def test_stop_after_failed_up_calls_cmd_stop(monkeypatch):
         return SimpleNamespace()
 
     monkeypatch.setattr("gpu_rent.cli.cmd_stop", fake_stop)
-    monkeypatch.setattr("gpu_rent.cli.warn", lambda *_a, **_k: None)
-    monkeypatch.setattr("gpu_rent.cli.ok", lambda *_a, **_k: None)
-    monkeypatch.setattr("gpu_rent.cli.err", lambda *_a, **_k: None)
-    monkeypatch.setattr("gpu_rent.cli.log", lambda *_a, **_k: None)
+    _quiet(monkeypatch)
     monkeypatch.setattr(
         "gpu_rent.state.load_state",
-        lambda: SimpleNamespace(floating_ip=None),
+        lambda: SimpleNamespace(server_id=None, floating_ip=None),
+    )
+
+    _stop_after_failed_up(SimpleNamespace(), RuntimeError("boom"))
+    assert calls == []
+
+
+def test_stop_after_failed_up_runs_when_server_exists(monkeypatch):
+    calls: list[dict] = []
+
+    def fake_stop(cfg, *, no_pull=False, log=None, destroy_disks=False):
+        calls.append({"no_pull": no_pull, "destroy_disks": destroy_disks})
+        return SimpleNamespace()
+
+    monkeypatch.setattr("gpu_rent.cli.cmd_stop", fake_stop)
+    _quiet(monkeypatch)
+    monkeypatch.setattr(
+        "gpu_rent.state.load_state",
+        lambda: SimpleNamespace(server_id="s1", floating_ip=None),
     )
 
     _stop_after_failed_up(SimpleNamespace(), RuntimeError("boom"))
@@ -33,20 +55,20 @@ def test_stop_after_failed_up_prints_digest_when_fip(monkeypatch):
         return SimpleNamespace()
 
     monkeypatch.setattr("gpu_rent.cli.cmd_stop", fake_stop)
-    monkeypatch.setattr("gpu_rent.cli.warn", lambda *_a, **_k: None)
-    monkeypatch.setattr("gpu_rent.cli.ok", lambda *_a, **_k: None)
-    monkeypatch.setattr("gpu_rent.cli.err", lambda *_a, **_k: None)
-    monkeypatch.setattr("gpu_rent.cli.log", lambda *_a, **_k: None)
+    _quiet(monkeypatch)
     monkeypatch.setattr(
         "gpu_rent.state.load_state",
-        lambda: SimpleNamespace(floating_ip="10.0.0.1"),
+        lambda: SimpleNamespace(server_id="s1", floating_ip="10.0.0.1"),
     )
     monkeypatch.setattr(
         "gpu_rent.vm_logs.print_log_digest",
         lambda cfg, host, **kw: digests.append(host),
     )
 
-    _stop_after_failed_up(SimpleNamespace(enable_swarmui=True, llm_runtime="none"), RuntimeError("boom"))
+    _stop_after_failed_up(
+        SimpleNamespace(enable_swarmui=True, llm_runtime="none"),
+        RuntimeError("boom"),
+    )
     assert digests == ["10.0.0.1"]
     assert calls == [{"no_pull": True}]