feat: add stub upstream when config is empty in dbless mode

[czeslavo]

What this PR does / why we need it:

When a translated configuration is empty, the controller adds an empty Upstream named kong to a Kong configuration. It will happen for Gateways in DB-less mode only.

Which issue this PR fixes:

Part of #3499.

PR Readiness Checklist:

Complete these before marking the PR as ready to review:

• the CHANGELOG.md release notes have been updated to reflect any significant (and particularly user-facing) changes introduced by this PR

[codecov]

Codecov Report

Patch coverage: 100.0% and project coverage change: -0.1 ⚠️

Comparison is base (5c4d7d5) 65.0% compared to head (891e03f) 65.0%.

Additional details and impacted files

@@           Coverage Diff           @@
##            main   #4316     +/-   ##
=======================================
- Coverage   65.0%   65.0%   -0.1%     
=======================================
  Files        154     154             
  Lines      17382   17384      +2     
=======================================
- Hits       11310   11301      -9     
- Misses      5353    5364     +11     
  Partials     719     719             

Impacted Files	Coverage Δ
internal/dataplane/deckgen/deckgen.go	52.7% <100.0%> (+4.9%)	⬆️
internal/dataplane/deckgen/generate.go	44.5% <100.0%> (+2.2%)	⬆️
internal/dataplane/kong_client.go	84.7% <100.0%> (-0.3%)	⬇️
...nal/dataplane/sendconfig/config_change_detector.go	100.0% <100.0%> (ø)

... and 2 files with indirect coverage changes

☔ View full report in Codecov by Sentry.
📢 Do you have feedback about the report comment? Let us know in this issue.

[rainest]

I'll always 🤨 a bit at variable renames without explanations why, but doesn't make any difference functionally. Gateway probably should report ready if it successfully receives an empty config and not require this stub config workaround, but assuming we probably can't get that, so

[pmalek]

Do you mind reminding me what's the reason for this? I do #3499 but as I see it now is that this will circumvent the /status/ready functionality: i.e. instead of waiting for first non-empty config we'll send the stub to make the GW ready which will serve basically nothing.

[czeslavo]

It's to make sure a Gateway always gets ready after getting the first (empty or not) config update from KIC. Effectively, Gateways using /status/ready will stay not-ready until KIC sends the first config update to them. IMO it's expected to have KIC configure Gateways with an empty config when there's nothing in K8s API to translate and to assume these Gateways will be ready to serve traffic in such case, isn't it?

This will also ensure that a typical scenario we run in tutorials etc. is not broken:

1. Deploy KIC + Kong using helm chart.
2. GET /path returns 404.

Without this workaround, we'd have no Kong instances ready to serve and respond 404s.

[pmalek]

👍