Spurious validation step when restarting with a checkpoint when `max_steps` is set in the trainer

Nobody has claimed this yet.

### Bug description

When restarting from an existing checkpoint with a trainer that has a `max_steps` value set, the trainer does a single validation step before actually restarting the training epoch (even if val_sanity_checks_step is set to 0)
This is an issue because that one validation step is considered a full validation round (callbacks are applied, which means the Tensorboard logger logs those spurious data points).
Note that this only applies when `max_steps` is set, and not `max_epochs` so there might be something going on with the management of the global step.

Also, this problem has been reported in this discussion:
https://github.com/Lightning-AI/lightning/discussions/18110

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

v2.0

### How to reproduce the bug

```python
import os
from torchvision.datasets import MNIST
import os
from torch import optim, nn, utils, Tensor
from torchvision.datasets import MNIST
from torchvision.transforms import ToTensor
import pytorch_lightning as pl
from torch.utils.data import DataLoader
import tempfile

# define any number of nn.Modules (or use your current ones)
encoder = nn.Sequential(nn.Linear(28 * 28, 64), nn.ReLU(), nn.Linear(64, 3))
decoder = nn.Sequential(nn.Linear(3, 64), nn.ReLU(), nn.Linear(64, 28 * 28))

# define the LightningModule
class LitAutoEncoder(pl.LightningModule):
def __init__(self, encoder, decoder):
super().__init__()
self.encoder = encoder
self.decoder = decoder

def training_step(self, batch, batch_idx):
# training_step defines the train loop.
# it is independent of forward
x, y = batch
x = x.view(x.size(0), -1)
z = self.encoder(x)
x_hat = self.decoder(z)
loss = nn.functional.mse_loss(x_hat, x)
# Logging to TensorBoard (if installed) by default
self.log("train_loss", loss)
return loss

def validation_step(self, batch, batch_idx):
x, y = batch
x = x.view(x.size(0), -1)
z = self.encoder(x)
x_hat = self.decoder(z)
loss = nn.functional.mse_loss(x_hat, x)
# Logging to TensorBoard (if installed) by default
self.log("val_loss", loss)
return loss

def on_validation_end(self) -> None:
print('Ending Validation')
return super().on_validation_end()

def configure_optimizers(self):
optimizer = optim.Adam(self.parameters(), lr=1e-3)
return optimizer

def setup_and_train(ckpt_path):
# init the autoencoder
autoencoder = LitAutoEncoder(encoder, decoder)

train_dataset = MNIST(os.getcwd(), download=True, transform=ToTensor())
test_dataset = MNIST(os.getcwd(),train=False, download=True, transform=ToTensor())
train_loader = utils.data.DataLoader(train_dataset)
test_loader = utils.data.DataLoader(test_dataset)

ckpt_dir = '/content/ckpt_dir'

checkpoint_callback = pl.callbacks.ModelCheckpoint(
dirpath=ckpt_dir,
save_weights_only=False,
save_top_k=0,
# monitor='EM',
# mode='max',
save_last=True,
)
# Initial training
trainer = pl.Trainer(limit_train_batches=2000, max_steps=100000, num_sanity_val_steps=0, val_check_interval=1.0, callbacks=checkpoint_callback)
trainer.fit(model=autoencoder, train_dataloaders=train_loader, val_dataloaders=test_loader, ckpt_path=ckpt_path)
```

I used this script changing the trainer configuration and stopping the run manually in a notebook. You can see that when restarting the run from a checkpoint with `max_steps` set, there's one val step run before resuming the training.
```

### Error messages and logs

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

### Environment

- toml: 0.10.2
- tomli: 2.0.1
- toolz: 0.12.0
- torch: 2.0.1+cu118
- torchaudio: 2.0.2+cu118
- torchdata: 0.6.1
- torchmetrics: 1.2.0
- torchsummary: 1.5.1
- torchtext: 0.15.2
- torchvision: 0.15.2+cu118
- tornado: 6.3.2
- tqdm: 4.66.1
- traitlets: 5.7.1
- traittypes: 0.2.1
- transformers: 4.33.2
- triton: 2.0.0
- tweepy: 4.13.0
- typer: 0.9.0
- types-setuptools: 68.2.0.0
- typing-extensions: 4.5.0
- tzlocal: 5.0.1
- uc-micro-py: 1.0.2
- uritemplate: 4.1.1
- urllib3: 2.0.4
- vega-datasets: 0.9.0
- wadllib: 1.3.6
- wasabi: 1.1.2
- wcwidth: 0.2.6
- webcolors: 1.13
- webencodings: 0.5.1
- websocket-client: 1.6.2
- werkzeug: 2.3.7
- wheel: 0.41.2
- widgetsnbextension: 3.6.5
- wordcloud: 1.9.2
- wrapt: 1.15.0
- xarray: 2023.7.0
- xarray-einstats: 0.6.0
- xgboost: 1.7.6
- xlrd: 2.0.1
- xxhash: 3.3.0
- xyzservices: 2023.7.0
- yarl: 1.9.2
- yellowbrick: 1.5
- yfinance: 0.2.28
- zict: 3.0.0
- zipp: 3.16.2
* System:
- OS: Linux
- architecture:
- 64bit
- ELF
- processor: x86_64
- python: 3.10.12
- release: 5.15.120+
- version: #1 SMP Wed Aug 30 11:19:59 UTC 2023

### More info

_No response_

cc @carmocca @justusschock

No source file or test is named. Start by running the provided checkpoint-resume reproduction with max_steps set and num_sanity_val_steps=0, then trace the trainer's resume and validation entry points. Done means restarting performs no validation step or callback before training resumes, so no spurious logged data points appear.

Written by the indexing model from the issue text.