Remove start.sh and switch to sg to run dev environment

[mrnugget]

2 weeks ago, in #24988, we announced that we'll remove start.sh in 2 weeks. Time's up.

This the majority of start.sh but leaves a message in it saying that we switched to sg and where to find documentation.

It also removes all the shell scripts that start.sh used to boot up the dev environment that we don't need anymore, either because they're built into sg (such as the watchman configuration, wrapper, helper scripts) or because they're now in sg.config.yaml (where we'll step-by-step reduce them by building more into sg).

[jeanduplessis]

🚀

[mrnugget]

Side note to @bobheadxi @jhchabran: see 8fd2d88. I had to restore dev/docsite.sh because we use docsite check in CI. That's fine for now, I think, but ideally we'd use sg in CI to run this kind of stuff too so it's easy to reproduce locally if something fails. Maybe we can do something like sg install-all && .bin/docsite check or sg run docsite -- new-args etc.

[jhchabran]

(...) ideally we'd use sg in CI to run this kind of stuff too so it's easy to reproduce locally if something fails. Maybe we can do something like sg install-all && .bin/docsite check or sg run docsite -- new-args etc.

I really like the idea and ideally I believe that the CI should be build on top of clearly identified interface that we use also locally as opposed to custom shell scripts that are made only for the CI (make it harder to maintain and it silos the knowledge). (RFC 484 broach that topic).

It totally ties sg to the main repo though, which is something we probably want to keep in mind.

[mrnugget]

It totally ties sg to the main repo though, which is something we probably want to keep in mind.

Yep, but I don't think it's bad. We do have a monorepo and there are disadvantages that come with pulling something out of the repository.