Fix encode_string function

[felixdittrich92]

Bug description

Currently there is no check if the single characters are also available in the given vocabulary.
We need a check for this :)

TODO's:

• check that in the function and throw a meaningful exception
• improve the corresponding test

discussion:
#926

Code snippet to reproduce the bug

from doctr.datasets.utils import encode_string
from doctr.datasets import VOCABS

x = encode_string(input_string='abcDÄÜ', vocab=VOCABS['english'])   # Ä and Ü does not exist in vocab
# raises ValueError: substring not found

Error traceback

Traceback (most recent call last):
  File "/home/felix/Desktop/doctr/test.py", line 7, in <module>
    x = encode_string(input_string='abcDÄÜ', vocab=VOCABS['english'])   # Ä and Ü does not exist in vocab
  File "/home/felix/Desktop/doctr/doctr/datasets/utils.py", line 75, in encode_string
    return list(map(vocab.index, input_string))  # type: ignore[arg-type]
ValueError: substring not found

Environment

not need :)

Deep Learning backend

same

[frgfm]

Thanks for the issue 🙏
I'd argue it's a documentation/error message clarity issue rather than a bug (meaning that we don't want the code to stop failing in this snippet)