Yugabyte: Working GCP implementation

[the-glu]

This PR introduce changes in the teraform files to support yugabyte in GCP.

It follow #1186 . You probably want to review the last commit only.

It needs the following PRs to make it working even with yugabyte not working correctly with server_broadcast_address ( yugabyte/yugabyte-db#27367 ), we do 'cheat' by injecting into the host file the public host we want, making yugabyte broadcasting the public hostname as private rpc address.

• Add more overrides for rpc bind address and server broadcast address yugabyte/charts#207
• Add pre commands options yugabyte/charts#208

It does the following:

• Added a new terraform option datastore_type, and needed yugabyte options
• Fixed and added a public name option in certificate generation for yugabyte nodes
  • The previous PR doesn't include the extension file, generating certificate only valid for the 'main' host
• Add correct load balances for yugabyte, exposing all ports
  • Some port are exposed even if not valid to avoid a UI issue doing queries on unavailable port and waiting for timeout (exposing them reject directly the tcp connection)

It doesn't have, for now (but the PR is quite big):

• Documentation
• Some variables are shared with cochroachdb and haven't been renamed to generic ones.
• AWS implementation is broken dues to changes in the common part, but next PR will fix that :)
• Default values

It has been tested with two independent GCP k8s clusters that successfully formed a cluster:

[barroco]

First pass of comments. May I kindly ask you to provide a list of ports required to be exposed in order to successfully pool two DSS instances?

[barroco]

Let's remove this and address it in another PR if needed. Since there is no autoscalling, this would only change the requested resources without changing the type of nodes provisionned.

[barroco]

Though, I would expect to see a configuration flag similar to https://github.com/interuss/dss/pull/1190/files#diff-f08dfea212cbd81c9b94266f15255cfd4c4b381b146e7b42c2049c77fbb5560eR24 to indicate the number of nodes.

[the-glu]

It's needed when testing on a small cluster (e.g. by following the default values / README) because the services are requesting a lot of resources, I would suggest to keep it to ease quick / test deployments without having the need of spawning a big cluster.

Not sure how it's related to autoscalling however ?

(Will do add a configuration flag for the number of nodes)

[the-glu]

(Number of nodes support added)

[barroco]

I don't see the definition of yugabyteTserverNodes. Is it required to expose publicly those ports ?

[the-glu]

Some are needed for inter-tserver communication, and some are needed for the interface / monitoring stuff.

Should we don't want to be able to use the UI, we could limit those, but then the UI will be in a half-working state.

[the-glu]

( yugabyteTserverNodes added )

[the-glu]

@barroco ready for a second pass

May I kindly ask you to provide a list of ports required to be exposed in order to successfully pool two DSS instances?

Where do you want that ? I suggest we do it in the cleanup/doc phase ( #1192 ) :)

[barroco]

Thanks, looks good to me. Let's merge it and address #1191 as quickly as possible to avoid inconsistency due to changes of terraform-commons-dss.
We can still do a pass of optimization before next release.