tensorboard step and self.global_step do not correspond under accumulate_grad

Nobody has claimed this yet.

### Bug description

Assume accumulate_grad=2, log_every_n_steps=50, val_check_interval=8000; tensorboard is the self.log. The self.global_step will add one when model.forward is done, but tensorboard step will add one when loss.step() is done. So my model will valid when tensorboard step = 4000, because this time the self.global_step is 8000.

I think this logic should be unified.

### What version are you seeing the problem on?

v2.4

### How to reproduce the bug

_No response_

### Error messages and logs

```
# Error messages and logs here please
```

### Environment

Current environment

```
#- PyTorch Lightning Version (e.g., 2.4.0):
#- PyTorch Version (e.g., 2.4):
#- Python version (e.g., 3.12):
#- OS (e.g., Linux):
#- CUDA/cuDNN version:
#- GPU models and configuration:
#- How you installed Lightning(`conda`, `pip`, source):
```

### More info

_No response_

cc @lantiga

Start by tracing how self.global_step and the TensorBoard step are updated around model.forward and loss.step when accumulate_grad is 2. Reproduce the reported mismatch with log_every_n_steps=50 and val_check_interval=8000, then determine the intended shared step semantics. Done means validation and TensorBoard use corresponding steps, with a regression test covering accumulation.

Written by the indexing model from the issue text.