Fixes Node Version pinning & updates with changes made in HCCE

[DougReeder]

What?

1. Fix Node Version pinning
2. Updates with changes made in hubs-cloud/community-edition/services/dialog/

Why?

1. the new lts of node is 20 now, so the docker build fails with incompatible node modules, because the build version is 18.x.x and the run version is 20.x.x
   Using the node version pinning also for the "runtime-version" fixes this issue.
   The other modifications are just to get rid of some docker build warnings.
2. It was necessary to include the changes made in HCCE to get the image to build
3. To consolidate development.
4. netcat was already in the base image
5. The healthcheck function was hacky and was less useful than fetching http://dialog:7000/meta

Examples

Kubernetes readiness & liveness probes should now use the endpoint http://dialog:7000/meta instead of http://dialog:1111/

How to test

1. Install image in Hubs instance:
   kubectl set image deployment/dialog dialog=dougreeder/dialog:match-ce-2026-01-23-20-03 -n hcce
   kubectl rollout status deploy dialog -n hcce
2. Check for regressions
3. Run kubectl top pod -n hcce and verify that dialog's memory usage is not more than 160MiB.

Documentation of functionality

No change in functionality.

Limitations

Does not do Mozilla branding removal

Alternatives considered

The commented-out code could have been left in, but there is no foreseeable reason it would ever be uncommented.

Additional details or related context

tanfarming commited the changes in hubs-cloud

Incorporates and supercedes #58 and #59

[Exairnous]

Thank you (and @B4dT0bi) for the PR. This builds and runs fine. I tried with four open tabs and the memory only rose to 50MB (after a total of 39 minutes of running it had only risen to 54MB).

I don't see any sign of http://dialog:7000/meta being implemented as a health check, though. hcce.yam in `hubs-cloud/community-edition/generate_script`` doesn't include any readiness or liveness probes.

I think this is what we planned, but I wanted to double check since the PR description suggests a new health check is part of this PR (or that it's falling back on a default health check).

Lastly, I think there's a typo in the Additional details or related context section of the PR description:
> Incorporates and supercedes #58 and #58

Could you update that, please?

[Exairnous]

Never mind the part about the typo fix, I saw the mentions of the two linked PRs that were closed and updated the description myself.

[DougReeder]

http://dialog:7000/meta is not specifically a health check, it's just a path that's part of the main server which returns a small payload that returns {"cap":42} where 42 appears to be the number of connected users.

hcce.yam doesn't include a health check for dialog, but we could add one. I wanted the PR that completely removes the old healthcheck endpoint to record what to use as a replacement.