docs: Refresh self install

[christopher-hakkaart]

No description provided.

[christopher-hakkaart]

@copilot resolve the merge conflicts in this pull request

[Copilot]

@copilot resolve the merge conflicts in this pull request

Merge conflicts resolved in docs/install/docker-compose.md and docs/install/reference.md in commit 1e4e10d.

[pditommaso]

Review: self-install docs refresh

The restructure is the right shape — two-path index, assemblable manifests, a shared verification page, settings pushed out to reference.md. But the pages don't run yet. Reviewed at 1e4e10d; every item below re-checked by hand against src/main.

Won't work

Two that stop a deployment cold:

1. aws-build.md:377 enables scan without wave.scan.reports.path. ObjectStorageOperationsFactory.create0 throws on a null path, and Application.groovy:43 starts Micronaut with eagerInitSingletons(true) — so Wave won't boot. reference.md:378 already marks the key required.
2. aws-build.md:272 creates wave-build-pvc in namespace wave, but line 362 runs build jobs in wave-build. PVCs are namespace-scoped, so every build/scan/mirror pod stays Pending.

And a regression: the merge commit reverted six keys in reference.md from kebab-case back to camelCase, undoing #1096 (master had all six right). They still bind from YAML because Micronaut hyphenates every key it loads, but WAVE_BLOBCACHE_STORAGE_ACCESSKEY no longer resolves — and lines 563/581 still say blob-cache.*, so the section contradicts itself.

Four more that fail silently:

• DOCKER_USER/DOCKER_PAT in wave.env (docker-compose.md:110) do nothing. Nothing in src/main reads those names — they work only via the config.yml the image ships (build.gradle:184), which this guide tells you to mount over.
• The MICRONAUT_ENVIRONMENTS strings in configure-wave.md:78,130,326 re-add lite, and WaveLite gates out JobManager, JobServiceImpl, both job queues and cleanup via @Requires(missingBeans = WaveLite).
• The endpoints.health guard master shipped (health.jdbc / health.disk-space off) was dropped, while /health is now the liveness probe in both guides. A brief Postgres blip restarts every replica.
• WAVE_JVM_OPTS was dropped from a 4 GiB pod; src/main/jib/launch.sh defaults to -Xmx850m.

Also: the registry snippet at kubernetes-lite.md:146 is a second top-level wave: block (Wave loads one YAML document); that page never says kubectl apply; its ingress has no ingressClassName or annotations so /service-info can't pass; the <s3-bucket> the ConfigMap depends on is never created; noarch is missing from the node selector so mirror pods skip the build node group; the production checklist uses the legacy wave.allowAnonymous alias instead of wave.capabilities.anonymous-access; and how-wave-works.md:126 still links ./configuration.md.

Simplify

Length is concentrated in three blocks: the seven-registry pre-creation matrix (aws-build.md:59-89 — on an EKS page), the IAM CLI narration (109-132, replaceable with one AWS link), and the SMTP table (configure-wave.md:99-109). Beyond that: Bottlerocket and "Production enhancements" belong in troubleshoot / the existing checklist, IRSA is explained three times across two files, and configure-wave.md documents an EC2 instance profile for Docker-based builds that index.md:23 says aren't supported.

A stacked PR against this branch follows with the fixes.