`configure_model` is incompatible with the `BaseFinetuning` behavior when fitting

Nobody has claimed this yet.

### Bug description

Based on the current callback orders, The `Finetuning` class will always be incompatible with any `LightningModule` that utilize `configure_model` method. The current callback sequence is
`Callback.setup`
-> `LightningModule.configure_model`
-> `LightningModule.configure_optimizers`
-> `Callback.on_fit_start`

However, The `BaseFinetuning` calls `freeze_before_training` at `setup`, where modules inside the `configure_model` has not been instantiated yet.

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

v2.1

### How to reproduce the bug

```python
from lightning import LightningModule
import torch
from torchvision import models
class MyModel(LightningModule):
def configure_model(self):
self.backbone = models.resnet18()
def configure_optimizers(self):
return torch.optim.SGD(lr=1e-3)
```

### Error messages and logs

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

### Environment

Current environment

```
#- Lightning Component (e.g. Trainer, LightningModule, LightningApp, LightningWork, LightningFlow):
#- PyTorch Lightning Version (e.g., 1.5.0):
#- Lightning App Version (e.g., 0.5.2):
#- PyTorch Version (e.g., 2.0):
#- Python version (e.g., 3.9):
#- OS (e.g., Linux):
#- CUDA/cuDNN version:
#- GPU models and configuration:
#- How you installed Lightning(`conda`, `pip`, source):
#- Running environment of LightningApp (e.g. local, cloud):
```

### More info

_No response_

Start by tracing the listed callback sequence around BaseFinetuning.setup, LightningModule.configure_model, configure_optimizers, and Callback.on_fit_start, then reproduce the issue with the provided MyModel example. Done means a LightningModule that creates modules in configure_model can be used with BaseFinetuning without the setup-time incompatibility.

Written by the indexing model from the issue text.