Description

This PR adds an initial locustfile for running load testing initially. It will need to be modified as the load testing plan develops but this gives you an idea of how load testing might work locally.

How to run the load test:
For now (eventually will make this a single script):

1. Spin up the API locally. I used ./scripts/local_server.sh
2. Run locust -f ./tests/load/locustfile.py --headless -u1 -r 1 --run-time 3s --host http://localhost:8000

Interpreting the output:
You will get two sets out output after the load test has completed:

1. of requests, # fails, and time for each endpoint that is hit

2. Approximate response time percentiles

We can save the output as a CSV for easier sharing with --csv example

Details about locustfile:

1. The seed data is loaded: I went with loading a simple file for now, but as we increase the number of records, we may need to read in the file incrementally in case it is too large to open in memory or generate the seed data on the fly.
2. On start, the seed data is added into the MPI: currently this runs once per user (not once per test) and will need to be fixed in future iterations. To avoid seeding multiple times, I have set the number of users to 1.
3. Randomize the seed data: A random record from the seed data is chosen and run through record linkage by hitting the /link endpoint. In future iterations, the record will be duplicated a random number of times and randomized once I have updated expand_test_data to work with PIIRecords (ticket Refactor expand_test_data.py to accept PIIRecords #376)

Related Issues

Closes #377

Additional Notes

There are likely lots of iterations we can make on this file as we learn more about how locust works so let me know if you find something neat!