White space at the start of each token_str. Could this affect backwards compatibility perhaps? Feature or Bug?

I just happened to realize that calls to BERT would output a `token_str = "word"` whereas a similar call to modernBERT outputs `token_str = " word"` with an additional white space before the `token_str`. Could this create some compatibility issues for people who want to swop in modernBERT to an existing workflow?

```
import torch
from transformers import pipeline
from pprint import pprint

input_text = "COVID is a [MASK]."

pipe = pipeline(
"fill-mask",
model="answerdotai/ModernBERT-base",
torch_dtype=torch.bfloat16,
)

results = pipe(input_text)
pprint(results)

pipe = pipeline(
"fill-mask",
model="bert-base-uncased",
torch_dtype=torch.bfloat16,
)
results = pipe(input_text)
pprint(results)
```

Reproduce the reported difference with the transformers fill-mask pipeline using ModernBERT-base and bert-base-uncased, then inspect how each model produces token_str. Determine whether the leading whitespace is intentional or a compatibility bug; done means the behavior is explained and any required change has a regression test or documented rationale.

Written by the indexing model from the issue text.