Some group_norm tests are failing in debug build

[bbradelTT]

Repro steps:

• build debug build
• run test_group_norm tests. Some testing showed that some subsets of the below may fail (checked on GS and WH)
• pytest tests/ttnn/unit_tests/operations/test_group_norm.py::test_group_norm_with_block_sharded_v2_8x8_grid
• pytest tests/ttnn/unit_tests/operations/test_group_norm.py::test_group_norm_with_block_sharded_v2_8x8_grid_tile_layout
• 'pytest tests/ttnn/unit_tests/operations/test_group_norm.py::test_group_norm_with_height_sharded'

Expected Behaviour:

• pass

Actual Behaviour:

• some tests fail

See tenstorrent/pytorch2.0_ttnn#538

Might be related to bad CB handling, e.g. see #15458

[bbradelTT]

On GS with Debug:

pytest "tests/ttnn/unit_tests/operations/test_group_norm.py::test_group_norm_with_block_sharded_v2_8x8_grid[N=2-C=320-H=64-W=64-num_groups=32-device_params={'l1_small_size': 0}]"

produces

E       RuntimeError: TT_ASSERT @ ../tt_metal/impl/buffers/circular_buffer_types.cpp:109: false
E       info:
E       Cannot set to globally allocated buffer. Circular buffer size 131072 B exceeds allocated L1 buffer bank size of 98304 B
...
E        --- tt::tt_metal::v0::CircularBufferConfig::set_globally_allocated_address(tt::tt_metal::v0::Buffer const&)
E        --- ttnn::operations::normalization::groupnorm_multi_core_sharded(tt::tt_metal::Tensor const&, std::__1::optional<tt::tt_metal::Tensor const> const&, std::__1::optional<tt::tt_metal::Tensor const> const&, std::__1::optional<tt::tt_metal::Tensor const> const&, tt::tt_metal::Tensor&, float, unsigned int, unsigned int, MathFidelity, tt::tt_metal::DataType, tt::umd::xy_pair, bool)