Move the webapp to use App Engine's flexible runtime

[jcscottiii]

Problem & Background

During the weekly deployment to production, the webapp can repeatably timeout. This can lead to lengthy deployments for the engineer on call. However, the search cache and results processor seem to have higher success rates. The leading theory behind this is because they both use App Engine's flexible runtime instead of the standard runtime.

• Processor (prod)(staging)
• Search cache (prod)(staging)

Proposal

Move the webapp service to use App Engine's flexible runtime

Steps:

• Make a new Dockerfile. Can be similar to Search Cache's Dockerfile.
  • We cannot reuse the same dockerfile because of this issue: https://issuetracker.google.com/issues/286782972
  • Can most likely just call make go_build
• Configure the staging and prod yamls to use the flexible runtime
• Take a look at this document for any gotchas
• Take a look at this document

Other added benefits

• We can debug the Dockerfile being built in GCP Cloud Build
  • Example Search Cache build
  • Example processor build

cc: @past

[past]

A puzzling observation is that the production deployment consistently fails whereas the staging deployment does not.

[past]

Completed via #4226.