Conformal bootstrap model

[lennybronner]

Description

This is a work in progress PR for the new bootstrap election model. This should hopefully start the code review process,

Things that still need happen:

1. here are a few changes in the data handlers that will be superseded by the Estimandizer and I'll make sure to replace the code.
2. continue adding unit tests
3. merging new OLS/QR from elex-solver
4. Finalize last model decisions

I would recommend reading the code to try and understand the new model alongside the technical walkthrough posted on the insights channel.

Jira Ticket

https://arcpublishing.atlassian.net/browse/ELEX-2677

Test Steps

Make sure you're on the corresponding branch in elex-solver, so in that repo do:

git pull
git checkout elex-2771-update-quantile-regression

Then in the elex-model repo after checking out this branch:

pip install {path to elex-solver}
elexmodel 2020-11-03_USA_G --office_id=P --estimands=margin --geographic_unit_type=county --pi_method=bootstrap --percent_reporting=30 --aggregates=postal_code --features=baseline_normalized_margin --fixed_effects=county_classification

This should also work from the bootstrap_model_testbed branch in the testbed repo.

[dmnapolitano]

Hi! 👋🏻 So, I'm sick 🤧 and don't have much functioning brain power, but I still wanted to give this a first pass and round of comments/questions. In general I think this is really cool 😄 🎉 but since it's so many changes, I'd like to look over it again, especially once some of my initial questions have been answered. Thanks! 🙌🏻

[dmnapolitano]

Some more questions 😄

[dmnapolitano]

LGTM!! 🎉