Adding another possibility to provide the num_words in order to compute this dynamically

[JJorgeDSIC]

In the context of creating the vocab using all the in-domain data plus the remaining words from other text sources, this job now can receive also a variable that comes from subtracting the in-domain words to the num_words originally specified.

i.e:

def get_vocab_prior_id(experiment_name: str, corpus_files: Dict[str, tk.Path], vocab_size: int, *kwargs) -> tk.Path:
    
    vocab_files_id = []
    corpus_files_od = {}
    for corpus_name, corpus_file in corpus_files.items():
        if "DOMAIN_TAG" in corpus_name:
            vocab_files_id.append(corpus_file)
        else:
            corpus_files_od[corpus_name] = corpus_file
    vocabulary_job = VocabularyFromTextJob(file_paths=vocab_files_id, num_words=vocab_size)
    voc_size_id = ComputeTextCorpusStatisticsJob(vocabulary_job.out_vocabulary).out_num_lines
    vocab_size_od = vocab_size - voc_size_id
    vocab_ood = get_vocab_combine_all(experiment_name, corpus_files_od, vocab_size_od)
    vocab_ood_id = ConcatenateJob([vocab_ood, vocabulary_job.out_vocabulary], zip_out=False, out_name="vocab").out
    return vocab_ood_id`