Jekyll Algolia Does Not Translate Markdown to HTML Resulting in Too Large Records

## I want to report a bug:

### What is the current behavior?
When processing my blog posts, sending posts to Algolia errors with `record too large` on a bunch of my records. Inspecting the log reveals failing records do not properly convert to HTML and are transmitting large chunks of markdown to algolia as the value of the HTML tag. Example value:

```

As a software developer, I spend almost every week at [Code for Boston](http://www.codeforboston.org). There are a large number of attendees consisting of optimistic boot camp graduates thirsty to get some experience coding on real open source projects. Having worked with over fifty boot camp graduates over the past year and a half and seeing their challenges, I have developed an understanding of the challenges boot camp graduates face as they transition to coding in the real world. While you can join me at Code for Boston to learn these things, I am going to share some of my wisdom below.\n\n### Git Skills\nIn order to effectively code in a team, you should have mastery of git and work with some kind of project management tool. For most of our open source projects, we use [GitHub](http://www.github.com) at MAPC and Code for Bos
```

### What is your expected behavior?
Based on [the documentation](https://community.algolia.com/jekyll-algolia/how-it-works.html#extracting-data) I would expect jekyll-algolia to translate my Markdown to HTML and then create single records for each paragraph.

### Git repository to reproduce the issue:
https://github.com/mzagaja/mzagaja.github.io is my blog.

### Ruby version used:
Ruby 3.1.2

### Jekyll version used:
jekyll 4.3.1

Reproduce the issue with the linked blog repository using Ruby 3.1.2, Jekyll 4.3.1, and the documented extracting-data behavior. Read the linked “How it works” documentation and trace the Markdown-to-HTML extraction path that creates Algolia records. Done means Markdown is converted to HTML and records are split by paragraph without triggering the record-size error.

Written by the indexing model from the issue text.