fix: address all code review issues

1. _merge_config: track explicit fields instead of comparing to defaults
2. plan_install: let asset resolution errors propagate (fail loudly)
3. _install_binary/_install_appimage: use argv lists instead of shell strings
4. _find_module: narrow exception to OSError/YAMLError, raise ConfigError
5. _install_binary: use pkg.extract_dir to scope binary search
6. plan_install: raise FlowError when pkg type needs PM but none found
7. Frozen dataclasses: change mutable list fields to tuples throughout
8. Remove dead stream_shell method and unused Console import
9. Guard os.getuid() with hasattr for cross-platform safety
10. _parse_targets: raise ConfigError on malformed entries
11. Bootstrap modules: use shlex.quote on all interpolated values

Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>

tests/test_domain_bootstrap_planning.py

@@ -26,7 +26,7 @@ class TestParseProfile:
         profile = parse_profile("minimal", {})
         assert profile.os == "linux"
         assert profile.hostname is None
-        assert profile.packages == []
+        assert profile.packages == ()
 
     def test_ssh_keys(self):
         raw = {"ssh-keys": [{"path": "~/.ssh/id_ed25519", "type": "ed25519"}]}