[BUG] reduce scatter cannot be overlap when using zero

Describe the bug
reduce scatter cannot be overlap when using zero

To Reproduce
DeepSpeed Configs:

json = {
    "train_batch_size": 64, 
    "train_micro_batch_size_per_gpu": 1, 
    "steps_per_print": 1, 
    "zero_optimization": {
        "stage": 3, 
        "overlap_comm": true
    }, 
    "gradient_clipping": 1.0, 
    "prescale_gradients": false, 
    "fp16": {
        "enabled": false, 
        "loss_scale": 0, 
        "loss_scale_window": 500, 
        "hysteresis": 2, 
        "min_loss_scale": 1, 
        "initial_scale_power": 11
    }, 
    "bf16": {
        "enabled": true
    }, 
    "wall_clock_breakdown": false
}

Expected behavior
reduce scatter should be overlap with computation

ds_report output

[2024-02-27 20:14:22,024] [INFO] [real_accelerator.py:191:get_accelerator] Setting ds_accelerator to cuda (auto detect)
--------------------------------------------------
DeepSpeed C++/CUDA extension op report
--------------------------------------------------
NOTE: Ops not installed will be just-in-time (JIT) compiled at
      runtime if needed. Op compatibility means that your system
      meet the required dependencies to JIT install the op.
--------------------------------------------------
JIT compiled ops requires ninja
ninja .................. [OKAY]
--------------------------------------------------
op name ................ installed .. compatible
--------------------------------------------------
 [WARNING]  async_io requires the dev libaio .so object and headers but these were not found.
 [WARNING]  async_io: please install the libaio-devel package with yum
 [WARNING]  If libaio is already installed (perhaps from source), try setting the CFLAGS and LDFLAGS environment variables to where it can be found.
async_io ............... [NO] ....... [NO]
fused_adam ............. [NO] ....... [OKAY]
cpu_adam ............... [NO] ....... [OKAY]
cpu_adagrad ............ [NO] ....... [OKAY]
cpu_lion ............... [NO] ....... [OKAY]
 [WARNING]  Please specify the CUTLASS repo directory as environment variable $CUTLASS_PATH
evoformer_attn ......... [NO] ....... [NO]
fused_lamb ............. [NO] ....... [OKAY]
fused_lion ............. [NO] ....... [OKAY]
inference_core_ops ..... [NO] ....... [OKAY]
cutlass_ops ............ [NO] ....... [OKAY]
transformer_inference .. [NO] ....... [OKAY]
quantizer .............. [NO] ....... [OKAY]
ragged_device_ops ...... [NO] ....... [OKAY]
ragged_ops ............. [NO] ....... [OKAY]
random_ltd ............. [NO] ....... [OKAY]
 [WARNING]  using untested triton version (2.0.0), only 1.0.0 is known to be compatible
sparse_attn ............ [NO] ....... [NO]
spatial_inference ...... [NO] ....... [OKAY]
transformer ............ [NO] ....... [OKAY]
stochastic_transformer . [NO] ....... [OKAY]
--------------------------------------------------
No CUDA runtime is found, using CUDA_HOME='cuda-11.7'
DeepSpeed general environment info:
torch install path ............... ['python3.10/site-packages/torch']
torch version .................... 1.13.1+cu117
deepspeed install path ........... ['deepspeed']
deepspeed info ................... 0.13.3+39f6ee59, 39f6ee59, master
torch cuda version ............... 11.7
torch hip version ................ None
nvcc version ..................... 11.7
deepspeed wheel compiled w. ...... torch 1.13, cuda 11.7

Screenshots

System info (please complete the following information):

• GPU count and types [4 machines with x8 A100s each]
• Interconnects [4 machines connected with 4IB]

No source file or test is named. Start by reproducing the supplied ZeRO Stage 3 configuration with overlap_comm enabled and trace the reduce-scatter path across the 4-machine, 8-A100-per-machine setup. Done means reduce-scatter overlaps computation in this bf16 configuration and a regression test covers the behavior.