refactor(prover): simplify to two Zelkova-like verification queries

Narrow OPP to two yes/no security questions:
1. Can data leave this sandbox? (exfiltration)
2. Can the agent write despite read-only intent? (write bypass)

Removes 4 advisory/lint queries (overpermissive methods, L4 gaps,
binary inheritance, inference relay), 3 output modes (Mermaid,
topology, HTML), and fixes 3 bugs (include_workdir not used in
exfil analysis, __all__ broken when protos missing, CLI PYTHONPATH).

~40% line reduction (2900 -> 1769 lines).

Signed-off-by: Alexander Watson <zredlined@gmail.com>

python/openshell/__init__.py

@@ -10,7 +10,7 @@ import contextlib
 # Proto stubs may not be generated yet (requires Rust build).
 # Suppress so subpackages like openshell.prover can still be imported.
 with contextlib.suppress(ImportError):
-    from .sandbox import (
+    from .sandbox import (  # noqa: F401 — intentional re-exports
         ClusterInferenceConfig,
         ExecChunk,
         ExecResult,
@@ -30,7 +30,7 @@ try:
 except Exception:
     __version__ = "0.0.0"
 
-__all__ = [
+_SDK_SYMBOLS = [
     "ClusterInferenceConfig",
     "ExecChunk",
     "ExecResult",
@@ -41,5 +41,6 @@ __all__ = [
     "SandboxRef",
     "SandboxSession",
     "TlsConfig",
-    "__version__",
 ]
+
+__all__ = [s for s in _SDK_SYMBOLS if s in dir()] + ["__version__"]