Switch from Elasticache to ECS Service for Redis

[willroberts]

Summary

Partial progress on #114.

While this increases costs in the short term (by about $2.33/mo), it will reduce costs in the long term when the AWS Free Tier expires, and ElastiCache starts to cost $11.68/mo. We can also use the second half of the new ECS instance to run our own Postgres database (at an additional cost of $1.04/mo for EBS disks which persist across containers).

Infrastructure changes (.github/, terraform/, etc.):

• Adds new modules for CloudMap (service discovery) namespaces and services
• Adds support for networking modes other than bridge in ECS services
• Adds support for service discovery in ECS services
• Adds support for running ECS services with images from outside our ECR registry
• Adjusts the default CPU amount from 1000 to 1024, since 2 cores provide 2048 CPU units and not 2000
• Enables DNS hostname resolution in the VPC
• Creates ECS services for Redis and Postgres
• Allows VPC traffic for Redis and Postgres
• Adds a CloudMap service for Redis
• Reconfigures the API, Game, SP, and Worker services to use the new ECS Redis service instead of ElastiCache
• Scales up the ECS cluster from 2 to 3 instances
• Cleans up ECS service module by removing unnecessary boolean variables (using existing variables instead)
• Shuts down the now-unused ElastiCache instance
• Updates docs

Testing

Have you have tested your changes in the following scenarios? Feel free to check off scenarios which don't apply.

• [x] Starting backend services locally with docker compose up succeeds.
• [x] I am able to create a new user and log in locally.
• [x] I am able to complete a practice game locally.
• [x] I am able to complete a purchase of Orbs, etc.