Add virtual memory support: PE_MMU, VA allocator, fabric MmuMapMsg

Implement VA/MMU layer (ADR-0011 Phase 1) enabling Triton kernels to use
contiguous virtual addresses on sharded tensors.

Key changes:
- PE_MMU component: hybrid inbox (MmuMapMsg) + sync translate() for PE_DMA
- VirtualAllocator + PEMemAllocator: free-list with coalescing
- MmuMapMsg/MmuUnmapMsg fabric path with SIP-level routing
- DPPolicy-based mapping: replicate=local, sharded=broadcast
- Tensor lifecycle: del + weakref cleanup, context manager
- Rename: TensorHandle.pa→addr, DmaReadCmd.src_pa→src_addr, ctx→torch

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

benches/loader.py

@@ -15,7 +15,7 @@ def resolve_bench(bench_id: str) -> BenchFn:
 
     Expected layout (repo root):
         benches/<bench_id>.py
-            def run(ctx: RuntimeContext) -> Any
+            def run(torch: RuntimeContext) -> Any
     """
     bench_id = bench_id.strip()
     if not bench_id:
@@ -30,7 +30,7 @@ def resolve_bench(bench_id: str) -> BenchFn:
 
     run_fn = getattr(mod, "run", None)
     if run_fn is None:
-        raise ValueError(f"Bench module {module_path} must define a 'run(ctx)' function.")
+        raise ValueError(f"Bench module {module_path} must define a 'run(torch)' function.")
     if not callable(run_fn):
         raise ValueError(f"'run' in {module_path} is not callable.")