Improve performance through client-side shaping

[jacobwegner]

As we continue to dig in to performance issues for EH, I've noticed that this type of response gets really heavy:

• get all of the text parts
  • for each text part, get all of the tokens
  • for each token, get all of the named entities

https://tinyurl.com/eh-perf-1

By splitting out the nested loop for named entities, we can cut the query time in half:

• get all of the text parts
  • for each text part, get all of the tokens
• get all of the named entities
  • for each named entities, get all of the tokens

https://tinyurl.com/eh-perf-2

There is still some shaping to be done on the queries (so we don't get tokens that aren't relevant to a particular passage, for example), but I like how this came out:

https://perf-data-shaping--explorehomer-dev.netlify.app/reader?urn=urn%3Acts%3AgreekLit%3Atlg0012.tlg001.perseus-eng4%3A1.1-1.160

vs

https://spike-apollo--explorehomer-dev.netlify.app/reader?urn=urn%3Acts%3AgreekLit%3Atlg0012.tlg001.perseus-eng4%3A1.1-1.160

and

https://perf-data-shaping--explorehomer-dev.netlify.app/reader?urn=urn%3Acts%3AgreekLit%3Atlg0012.tlg001.perseus-eng4%3A2.480-2.840

vs

https://spike-apollo--explorehomer-dev.netlify.app/reader?urn=urn%3Acts%3AgreekLit%3Atlg0012.tlg001.perseus-eng4%3A2.480-2.840

[jacobwegner]

Reduce nested queries

[jacobwegner]

closed in favor of #57