Add title topic modelling

[lyashevska]

Relates to ptypes-nlesc/ptypes-topic-modeling#14

Text Cleaning

1. Split the title into individual words.
2. Remove stopwords from the list of words.
3. Apply a word rooter function to each word, unless the word contains a '#'.
4. If the bigrams option is enabled, generate bigrams from the list of words and add them to the list.

Text Vectorization

The text vectorization process has been updated with the following changes:

1. A CountVectorizer is used to transform the cleaned titles into a frequency matrix.
2. The max_df parameter is set to 0.9, discarding words that appear in more than 90% of the titles.
3. The min_df parameter is set to 25, discarding words that appear in less than 25 titles.
4. The token_pattern parameter is set to '\w+|$[\d.]+|\S+', which matches words, dollar amounts, and non-whitespace sequences.

The frequency matrix is converted to an array and the feature names are retrieved using the get_feature_names_out method.