fix: apply all PR review agent findings (silent failures, type safety, test gaps)

Addresses issues found by 4 parallel review agents (code-reviewer,
silent-failure-hunter, type-design-analyzer, pr-test-analyzer).

Source fixes:
- core/utils.py: add public safe_display_url() (moved from tools/health.py)
- core/client.py: rename _redact_sensitive → redact_sensitive (public API)
- core/types.py: add SubscriptionData.__post_init__ for tz-aware datetime
  enforcement; remove 6 unused type aliases (SystemHealth, APIResponse, etc.)
- subscriptions/manager.py: add exc_info=True to both except-Exception blocks;
  add except ValueError break-on-config-error before retry loop; import
  redact_sensitive by new public name
- subscriptions/resources.py: re-raise in autostart_subscriptions() so
  ensure_subscriptions_started() doesn't permanently set _subscriptions_started
- subscriptions/diagnostics.py: except ToolError: raise before broad except;
  use safe_display_url() instead of raw URL slice
- tools/health.py: move _safe_display_url to core/utils; add exc_info=True;
  raise ToolError (not return dict) on ImportError
- tools/info.py: use get_args(INFO_ACTIONS) instead of INFO_ACTIONS.__args__
- tools/{array,docker,keys,notifications,rclone,storage,virtualization}.py:
  add Literal-vs-ALL_ACTIONS sync check at import time

Test fixes:
- test_health.py: import safe_display_url from core.utils; update
  test_diagnose_import_error_internal to expect ToolError (not error dict)
- test_storage.py: add 3 safe_get tests for zero/False/empty-string values
- test_subscription_manager.py: add TestCapLogContentSingleMassiveLine (2 tests)
- test_client.py: rename _redact_sensitive → redact_sensitive; add tests for
  new sensitive keys and is_cacheable explicit-keyword form

tests/test_info.py

@@ -186,7 +186,7 @@ class TestUnraidInfoTool:
     async def test_generic_exception_wraps(self, _mock_graphql: AsyncMock) -> None:
         _mock_graphql.side_effect = RuntimeError("unexpected")
         tool_fn = _make_tool()
-        with pytest.raises(ToolError, match="unexpected"):
+        with pytest.raises(ToolError, match="Failed to execute info/online"):
             await tool_fn(action="online")
 
     async def test_metrics(self, _mock_graphql: AsyncMock) -> None:
@@ -201,6 +201,7 @@ class TestUnraidInfoTool:
         _mock_graphql.return_value = {"services": [{"name": "docker", "state": "running"}]}
         tool_fn = _make_tool()
         result = await tool_fn(action="services")
+        assert "services" in result
         assert len(result["services"]) == 1
         assert result["services"][0]["name"] == "docker"
 
@@ -225,6 +226,7 @@ class TestUnraidInfoTool:
         }
         tool_fn = _make_tool()
         result = await tool_fn(action="servers")
+        assert "servers" in result
         assert len(result["servers"]) == 1
         assert result["servers"][0]["name"] == "tower"
 
@@ -248,6 +250,7 @@ class TestUnraidInfoTool:
         }
         tool_fn = _make_tool()
         result = await tool_fn(action="ups_devices")
+        assert "ups_devices" in result
         assert len(result["ups_devices"]) == 1
         assert result["ups_devices"][0]["model"] == "APC"