test(stateful): bound how far append grows an axis in the array lifecycle test (#4470)

`append` could add up to 4 rows every step with no upper bound, so over the
nightly profile's 100 steps an array with chunks of 1 reached tens of thousands
of chunks and each `matches_model` read took seconds. The 500 nightly examples
together exceeded pytest's 600s faulthandler_timeout (563s on the last green
nightly, killed on 2026-10-03). Cap appended extents at 2 * MAX_SIDE.

Assisted-by: ClaudeCode:claude-opus-5-5

Co-authored-by: Claude Opus 5.5 <noreply@anthropic.com>

Author: d-v-b

tests/test_array_stateful.py

@@ -50,6 +50,10 @@
 DTYPE = np.dtype("int16")
 METADATA_KEYS = (".zarray", ".zattrs", "zarr.json")
 MAX_SIDE = 6
+# `append` may grow an axis past `MAX_SIDE`, but not without bound: with chunks of 1,
+# the cells an extent of 100 steps could reach make each full-array read take seconds,
+# and the nightly run of every example together exceeded pytest's faulthandler_timeout.
+MAX_APPEND_EXTENT = 2 * MAX_SIDE
 
 
 async def _list(store: MemoryStore, prefix: str) -> list[str]:
@@ -210,7 +214,8 @@ def append(self, data: st.DataObject) -> None:
             axes = st.sampled_from(self.legacy_axes) | axes
         axis = data.draw(axes, label="axis")
         block_shape = list(self.shape)
-        block_shape[axis] = data.draw(st.integers(0, 4), label="rows")
+        max_rows = min(4, max(0, MAX_APPEND_EXTENT - self.shape[axis]))
+        block_shape[axis] = data.draw(st.integers(0, max_rows), label="rows")
         block = data.draw(npst.arrays(DTYPE, tuple(block_shape)), label="block")
         note(f"append {block.shape} along {axis} to {self.shape}")
         if self.shape[axis] == 0 and block.shape[axis]: