Fix typos chapter 6

[AlessandroMiola]

Hi everybody :)
first off, thank you for the amazing course and for your work on the HF ecosystem!

Here are a couple of fixes related to chapter 6 that I hope could be helpful:

• 6.2: simple typos
• 6.3: sequence_ids() (rather than sentence_ids()) helps mapping each token to the sentence it comes from (perhaps it was called sentence_ids() in previous versions?); then I'm adding the index keys to result so as to align with the provided output
• 6.5: substitution of attribute _tokenizer with backend_tokenizer (despite being the same) for coherence with 6.4 where backend_tokenizer is referenced
• 6.6: there seems to be an issue in the computation of WordPiece scores of token pairs ('hu', '##g') and ('hu', '##gs'); score for ('hu', '##g') should rather be 2/45 and this also affects the tokenization of word 'hugs' which is referred to in the next section (also referenced in Potential calculation error on WordPiece section #500)
• 6.7: there seems to be an issue in the computation of the unigram probabilities of word 'pug' (also referenced in https://discuss.huggingface.co/t/chapter-6-questions/11745/30).

[HuggingFaceDocBuilderDev]

The docs for this PR live here. All of your documentation changes will be reflected on that endpoint.