[Bug]: Problematic character removal does not correct for position in source text.

[chriskamphuis]

Describe the bug

Some problematic character are removed before creating a Sentence object:

flair/flair/data.py

Line 1108 in 7eb8533

def __remove_zero_width_characters(text: str) -> str:

When this happens, the location to the source text is not preserved. Perhaps these should only be removed after creating the tokens, such that position to source text is preserved.

To Reproduce

from flair.models import SequenceTagger
from flair.data import Sentence

tagger = SequenceTagger.load("flair/ner-english-large") 

text = "Hello my name is \u200c Chris Kamphuis, and I live in \u200c the Netherlands."
sentence = Sentence(text)
tagger.predict(sentence)
for span in sentence.get_spans():
    print(text[span.start_position:span.end_position])

produces:

 Chris Kamphui
e Netherlan

Expected behavior

It should produce:

Chris Kamphuis
the Netherlands

### Logs and Stack traces

_No response_

### Screenshots

_No response_

### Additional Context

_No response_

### Environment

#### Versions:
##### Flair
0.14.0
##### Pytorch
2.5.1+cu124
##### Transformers
4.44.2
#### GPU
False