Preserve backwards compatibility of teleport.lib.teleterm.v1.Gateway

[ravicious]

#26441 introduced what is technically a breaking change in teleterm's protos by changing the type of a field while reusing its name and number. This tripped up the CI check that Edoardo is working on (#26441 (comment)).

This PR corrects this by reserving the old field name and number and adding a new field with the new type.

[ravicious]

Oops, I forgot to change the Go code. 🤦

[ravicious]

For the Go code I changed only what's necessary to make the code compile. The Go code talks about "CLI command" much more, which kinda makes sense in that case since it operates on os.exec.Cmd. While "CLI command" as a name in general makes less sense (command-line interface command), I think I'll just leave it like that for now.

[ibeckermayer]

Granted that I'm not up to speed on what's going on in this section of the code, but at first glance a GatewayCLIClient that only sends a single command makes less sense than a GatewayCLICommand. I typically think of a client as a thing that sends and receives messages repeatedly, whereas afaict this object represents just a single command sent once.

I don't quite understand what you think doesn't make sense about "command-line interface command", to me that sounds like "a command that is sent to the command-line interface".

[espadolini]

(not entering the merit of how to call the field and the message)

[ravicious]

I hear your concerns @ibeckermayer. It is a single command that's supposed to launch the client. I'm going to use Edoardo's suggestion, gateway_cli_command, which I think addresses your concerns and at the same time doesn't necessitate as many changes in the frontend code as cli_client does.

[ravicious]

Oops, I forgot that I should probably wait for all reviewers before enabling auto merge, I hope all is good @ibeckermayer.