Fix Silo-HDF5 reader to parse Named Datatype structure correctly

The silo_reader was looking for coordinate/variable data as HDF5 Groups
and Datasets, but MFC's Silo files store objects as HDF5 Named Datatypes
with a compound "silo" attribute containing metadata (mesh name, data
paths, dimensions). Actual data arrays live under the .silo/ group.

Rewrite the reader to:
- Find mesh by silo_type=130 (DB_QUADMESH) on Named Datatypes
- Find variables by silo_type=501 (DB_QUADVAR) on Named Datatypes
- Resolve coord0/coord1/value0 paths from silo attribute to .silo/ datasets
- Fix timestep discovery to match actual file naming (<step>.silo)
- Clean up multi-processor assembly logic

Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>

Author: sbryngelson

toolchain/mfc/viz/reader.py

@@ -229,11 +229,11 @@ def discover_timesteps(case_dir: str, fmt: str) -> List[int]:
         p0_dir = os.path.join(case_dir, 'silo_hdf5', 'p0')
         if os.path.isdir(p0_dir):
             steps = set()
-            for dname in os.listdir(p0_dir):
-                if dname.startswith('t_step='):
+            for fname in os.listdir(p0_dir):
+                if fname.endswith('.silo') and not fname.startswith('collection'):
                     try:
-                        steps.add(int(dname.split('=')[1]))
-                    except (ValueError, IndexError):
+                        steps.add(int(fname[:-5]))
+                    except ValueError:
                         pass
             return sorted(steps)