Re-adds the alias so the Replicated SDK deployment is named
gameshelf-sdk as required by rubric 2.1. Also updates all
references (SDK_SERVICE_URL, support-bundle selectors/analyzers,
values key, helmchart.yaml key) to match.

This intentionally reverts the alias removal from fix/sdk-alias
for the Helm install demo path. The EC license injection issue
is deferred to a separate track.

Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>

Replaces proxy.replicated.com with the configured custom domain
so all pod images satisfy rubric 2.2 (images must start with
your custom proxy domain, not the generic proxy.replicated.com).

Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>

- helmchart.yaml: use integration.licenseID / integration.enabled
  (correct SDK subchart value path, not integrationLicenseID)
- README: add Helm Install section with full install/upgrade commands,
  pull secret setup, and common value overrides

Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>

Set gameshelf-sdk.image.registry to proxy.adamanthony.dev so the
SDK pod image also starts with the custom proxy domain. Without
this the SDK pulls directly from proxy.replicated.com, failing
the 2.2 requirement that every app image uses the custom proxy.

Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>

The logs collector writes to <pod-name>/<container-name>.log,
not <collectorName>/<pod-name>.log. The gameshelf app pod logs
land at e.g. gameshelf-66b5bdfddc-8d52j/gameshelf.log, so the
glob gameshelf-app/*.log never matched anything.

Use */gameshelf.log to match the container name regardless of
pod name hash.

Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>

Adds POST /admin/support-bundle route that collects lightweight app
diagnostics (app-info.json, license.json), packs them into a tar.gz
in memory using stdlib, and uploads to the Vendor Portal via the
Replicated SDK sidecar (POST /api/v1/supportbundle). Result (bundle ID
or error) is shown inline on the admin panel.

Note: uses an in-memory diagnostic bundle rather than kubectl
support-bundle to avoid adding the troubleshoot library (~267 deps,
50-100MB binary impact) or requiring the binary in the container image.
In production, collect a real bundle and POST the archive to the SDK.

Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>

runPod collector with collectorName db-connection-check writes to
db-connection-check.log at the root, not db-connection-check/check.log.

Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>

Expands the Helm Install section with:
- Full install command including SDK proxy, NodePort, integration mode
- Known gotchas: image tag mismatch, pull secret, SDK integration mode
  value path, SDK image proxy override
- Preflight demo commands: fail scenario via helm template, pass scenario
- Support bundle: laptop vs in-cluster approaches, RBAC setup,
  deploymentStatus failure induction (3.4), DB error induction (3.5)

Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>