[ngram] Improve tokenization regarding punctuation, @mentions, hashtags

[andi-halim]

Background
Default ngram analyzer removes punctuation for the default analyzer. Furthermore, there should be code written out to do the same for @mentions and hashtags, but this will be implemented later on as an n-gram customization

Problems
Punctuation, @mentions and hashtags are quite dominant in certain ngram analysis output, we need to allow this to be customizable.

Desired Outcome
Not splitting tokens based on punctuation.

Tasks

• Punctuation functionality
• Punctuation implemented in default ngram analyzer
• @mention filtering functionality
• hashtag filtering functionality