Remove dead differ-state-reporter stack and its workflow

[gsavage]

Removes the nested deployment/terraform/environment-reporter/ stack and the
tf-environment-reporter.yml workflow that applied it.

The stack deployed a lambda, differ-state-reporter, meant to report differ's Terraform
state file to a Kosli environment named terraform-state-differ-<env> on a rate(1 minute)
cron. It had been broken since ~2025-04: every invocation crashed at init with
Runtime.InvalidEntrypoint, so it never reported anything and its target Kosli environments
were never created. It is superseded by a consolidated state-file reporter that covers every
service at once, and differ was the only service still carrying a per-service reporter of
this kind.

The live AWS resources were destroyed by hand first, in both accounts — beta
244531986313 and prod 274425519734, 16 resources each (the lambda and its permissions,
two EventBridge rules and their targets, the IAM role, two policies and their attachments,
and the log group). Neither destroy touched anything shared: the state buckets and the
kosli_api_token SSM parameter are read-only inputs to the stack, and the stack's own state
lives under a separate key from the differ app stack. So this PR removes configuration for
infrastructure that no longer exists, rather than orphaning anything.

The directory and the workflow go together deliberately. The workflow triggers on pushes
under the stack's path with tf_apply: 'true', so deleting the directory on its own would
fire a run against a working_directory that no longer exists.

docs/dead-state-reporter.md is marked completed rather than deleted, so the full
investigation record stays in the repo.

🤖 Generated with Claude Code