fixtures: default batch_size to 2048 (DuckDB STANDARD_VECTOR_SIZE)

CountBatchArgs.batch_size and BATCH_SIZE_FALLBACK defaulted to 1000. Align the
canonical batch-size default with DuckDB's default vector size (2048) so the
standard sequence-family generators (sequence, double_sequence, nested_sequence,
profiling_demo) match the Rust/TypeScript workers, which already default to
2048. Deliberately-varied per-fixture batch sizes are unchanged.

Behaviorally invisible to output (batch_size only controls chunking); the one
shared SLT that asserts a batch count (dynamic_to_string.test) pins batch_size
explicitly, so cross-language workers stay consistent.

Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>

Author: rustyconover

vgi/_test_fixtures/table/_common.py

@@ -61,7 +61,7 @@ class CountBatchArgs:
     """
 
     count: Annotated[int, Arg(0, doc="Number of rows to generate", ge=0)]
-    batch_size: Annotated[int, Arg("batch_size", default=1000, doc="Batch size for output", ge=1)]
+    batch_size: Annotated[int, Arg("batch_size", default=2048, doc="Batch size for output", ge=1)]
 
 
 @dataclass(slots=True, frozen=True)
@@ -91,7 +91,7 @@ class _BaseSequenceFunction(TableFunctionGenerator[Any, CountdownState]):
     NUMPY_DTYPE: ClassVar[type[np.generic]] = np.int64
     STATS_ARROW_TYPE: ClassVar[pa.DataType] = pa.int64()
     STATS_COLUMN_NAME: ClassVar[str] = "n"
-    BATCH_SIZE_FALLBACK: ClassVar[int] = 1000
+    BATCH_SIZE_FALLBACK: ClassVar[int] = 2048
 
     @classmethod
     def initial_state(cls, params: ProcessParams[Any]) -> CountdownState: