Circular buffers goes out of memory if num cores are increased in unet model

[nkpatel-tt]

if we increase number of cores for upsample, having upsample changes from PR, circular buffers are going out of memory.
Following is replaced determine_num_cores_for_upsample

def determine_num_cores_for_upsample(nhw: int, width: int, max_cores=64) -> int:
    max_nshards = min(nhw, max_cores)
    nshards = max_nshards
    while nshards > 0:
        if nhw % nshards == 0:
            break
        nshards -= 1
    return nshards

After further debug, It seems that following configuration is failing for x = self.upblock3(x, c2_residual) for unet_shallow_ttnn.py.

y.shape: ttnn.Shape([1, 1, 42240, 96])
y.memory_config: MemoryConfig(memory_layout=TensorMemoryLayout::HEIGHT_SHARDED,buffer_type=BufferType::L1,shard_spec=ShardSpec(grid={[(x=0,y=0) - (x=7,y=7)]},shape={660, 96},orientation=ShardOrientation::ROW_MAJOR,halo=0,mode=ShardMode::PHYSICAL,physical_shard_shape=std::nullopt))
y.is_sharded: True
y.shard_spec: ShardSpec(grid={[(x=0,y=0) - (x=7,y=7)]},shape={660, 96},orientation=ShardOrientation::ROW_MAJOR,halo=0,mode=ShardMode::PHYSICAL,physical_shard_shape=std::nullopt)

It seems this particular configuration needs reshard after introducing above new determine_num_cores_for_upsample which is likely to cause issue.