Container Size ADR

We need to create an ADR to suggest a recommended max container size, tracking issues which we've run into while attempting to get very large containers through the pipeline.

Example projects:

• https://repo1.dso.mil/dsop/mathworks/matlab/matlab-with-desktop
• https://repo1.dso.mil/dsop/opensource/triton-inference-server

These containers will not be able to pass through the pipeline, so we'll need to come up with a reasonable recommendation for max container size to Nic.

Information needed for this document:

• Each tool's (or job's) maximum possible container/artifact size
  • Might need to define this more explicitly for artifacts. What is the max size for individual artifacts and all artifacts combined?
• Each tool's max image layer size.
• Any ops side resource limits that we can provide (probably just storage/memory for each runner but maybe include more info).
• Discuss methods for shrinking artifacts and layers
• What is our soft limit vs. hard limit and conclusions for why these limits are being recommended.