[WASM] SChunk.__getitem__/get_slice raises RuntimeError on Pyodide 0.29.4 (works on 0.29.3 and 314.0.0)

## Summary
On the WASM/Pyodide build, slicing an **in-memory** `SChunk` that was loaded from
a frame fails with `RuntimeError: Error while getting the slice` — but **only
under Pyodide 0.29.4**. The same code passes on Pyodide **0.29.3** (cp313) and
**314.0.0** (cp314).

The wheel tag is `pyemscripten_2025_0_wasm32` (ABI-level), so a wheel built/tested
against 0.29.3 still runs on a user's 0.29.4 runtime — meaning this is a genuine
runtime failure for 0.29.4 users, not just a CI artifact.

## Reproducer
`tests/test_open.py::test_load_schunk_returns_in_memory_copy`:

```python
urlpath = tmp_path / "schunk.b2frame"
data = np.arange(20, dtype=np.int32)
blosc2.SChunk(data=data, urlpath=urlpath, mode="w",
cparams={"typesize": data.dtype.itemsize})
loaded = blosc2.load(urlpath) # in-memory copy, urlpath is None
assert loaded[:] == data.tobytes() # <-- RuntimeError here
```

## Traceback

schunk.py:1114 __getitem__ -> get_slice(item.start, item.stop)
schunk.py:1058 get_slice -> super().get_slice(start, stop, out)
blosc2_ext.pyx:2058 -> RuntimeError: Error while getting the slice

## Environment matrix

```
│ Python / ABI │ Pyodide │ Result │
│ cp313 / pyemscripten_2025_0 │ 0.29.3 │ ✅ pass │
│ cp313 / pyemscripten_2025_0 │ 0.29.4 │ ❌ RuntimeError │
│ cp314 / pyemscripten_2026_0 │ 314.0.0 │ ✅ pass │
```

## Current workaround

.github/workflows/cibuildwheels.yml pins CIBW_PYODIDE_VERSION: 0.29.3 for the
cp313 wasm job (cibuildwheel 4.1's default is 0.29.4). This keeps CI green and
matches the wheels wasm.yml already ships, but does not fix the runtime for
0.29.4 users.