add train pipeline class in logging (meta-pytorch#3486)

Summary:
Pull Request resolved: meta-pytorch#3486

# context
* add TrainPipeline class logging to help debugging which train pipeline is used in the training job
* this will produce the following log
* the logging is done once when the train pipeline class is initiated (per rank)
```
INFO:torchrec.distributed.train_pipeline.train_pipelines:TrainPipeline class: <class 'torchrec.distributed.train_pipeline.train_pipelines.TrainPipelineSparseDist'>
```

Reviewed By: nipung90

Differential Revision: D85577739

fbshipit-source-id: 710a84cdb611cbaa2d255e40fb8464ee3826dac3

Author: TroyGarden

torchrec/distributed/train_pipeline/train_pipelines.py

@@ -110,6 +110,8 @@ def __init__(self) -> None:
         # pipeline state such as in foward, in backward etc, used in training recover scenarios
         self._state: PipelineState = PipelineState.IDLE
 
+        logger.info(f"TrainPipeline class: {type(self)}")
+
     def sync_embeddings(
         self,
         model: torch.nn.Module,