Document AWS CLI token expiry impact on CDK deployments

[scottgerring]

Problem

AWS deployments can take 30+ minutes across multiple CDK stacks. Organizations using temporary AWS credentials (SSO, STS assume-role) can hit token expiry mid-deploy, causing partial failures.

If a user panics and manually deletes CloudFormation stacks, this orphans CDK's local context (cdk.context.json), causing subsequent deploys to fail with confusing errors.

Suggested Fix

1. Each per-service deploy task (e.g. //sticker-award:aws:deploy) should run aws sts get-caller-identity before starting and produce a clear, helpful error message if credentials are invalid or expired — rather than failing deep into a CDK deploy.

2. Add a troubleshooting/prerequisites section to the deployment docs covering:

   • AWS deployments can take 30+ minutes — ensure you have a fresh token before starting
   • If interrupted by token expiry: re-authenticate and re-run mise run aws:deploy:local. CDK/CloudFormation will resume from where it left off
   • Do not manually delete CloudFormation stacks. If you already did, delete cdk.context.json in the relevant infra/aws directories before re-deploying

Part of #235