Modify Sentence.to_original_text() to take into account Sentence.start_position for whitespace calculation

[mauryaland]

Related to #3142. This solution, by taking the difference between the first token start index and the sentence start index, keep the initial whitespace for some special cases while not printing a lot of useless whitespaces because we store the index of the token from a bigger document.

For example for the following text which is split in two Sentence objects in my application:

text = "Amaury et Valentin mangent au 77 boulevard  Perreire.\n\n\n\n Paul  a modifié le tokenizer."

It gives before the fix:

[Sentence[9]: "Amaury et Valentin mangent au 77 boulevard  Perreire.",
 Sentence[6]: "                                                          Paul  a modifié le tokenizer."]

And after the fix:

[Sentence[9]: "Amaury et Valentin mangent au 77 boulevard  Perreire.",
 Sentence[6]: "Paul  a modifié le tokenizer."]

And a sentence that begins with a whitespace

sentence = Sentence(" ... and then?")

print(sentence)

still got it in the printout:

Sentence[4]: " ... and then?"

[alanakbik]

@mauryaland thanks for fixing this!