Create a benchmark get_attr_one_obj_per_big_alloc.

The behavior is currently quadratic that shouldn't be a case.

Benchmark                                          Time             CPU   Iterations
------------------------------------------------------------------------------------
get_attr_one_obj_per_big_alloc/size:100        16264 ns        16261 ns        36290
get_attr_one_obj_per_big_alloc/size:512       264089 ns       264055 ns         2615
get_attr_one_obj_per_big_alloc/size:4096    17013654 ns     17012206 ns           43
get_attr_one_obj_per_big_alloc/size:10000   93770372 ns     93755818 ns            7

PiperOrigin-RevId: 718795326
Change-Id: I0f98c7c57c70014ea0b9320a39ab5656139f0ccf

Author: goldvitaly

py/koladata/benchmarks.py

@@ -266,6 +266,22 @@ def get_attr_native(state):
     _ = ds.abc
 
 
+@google_benchmark.register
+@google_benchmark.option.arg_names(['size'])
+@google_benchmark.option.range(100, 10000)
+def get_attr_one_obj_per_big_alloc(state):
+  size = state.range(0)
+  alloc_size = 17
+  objs = []
+  schema = kd.uu_schema(a=kd.FLOAT32)
+  for i in range(size):
+    obj = schema(a=kd.slice([i] * alloc_size))
+    objs.append(obj.S[0])
+  ds = kd.slice(objs)
+  while state:
+    _ = ds.a
+
+
 @google_benchmark.register
 def uuid_item(state):
   a = kd.slice(42)