Refactor Genesis Code

[666lcz]

It is confusing to see the usage of genesis_ctx and pre_loaded_modules in both /sui/genesis path and sui/start path. This PR remove the usage of genesis related code from the /sui/start code path.

sui/sui/src/rest_server.rs

Lines 279 to 325 in 13c4e9f

	async fn sui_start(
	rqctx: Arc<RequestContext<ServerContext>>,
	) -> Result<HttpResponseOk<String>, HttpError> {
	let server_context = rqctx.context();
	let network_config_path = &server_context.network_config_path;
	let network_config = NetworkConfig::read_or_create(&PathBuf::from(network_config_path))
	.map_err(|error| {
	custom_http_error(
	StatusCode::CONFLICT,
	format!("Unable to read network config: {error}"),
	)
	})?;
	if network_config.authorities.is_empty() {
	return Err(custom_http_error(
	StatusCode::CONFLICT,
	String::from("No authority configured for the network, please make a POST request to the `sui/genesis` endpoint."),
	));
	}
	{
	if !(*server_context.authority_handles.lock().unwrap()).is_empty() {
	return Err(custom_http_error(
	StatusCode::FORBIDDEN,
	String::from("Sui network is already running."),
	));
	}
	}
	let committee = Committee::new(
	network_config
	.authorities
	.iter()
	.map(|info| (*info.key_pair.public_key_bytes(), info.stake))
	.collect(),
	);
	let mut handles = FuturesUnordered::new();
	for authority in &network_config.authorities {
	let server = sui_commands::make_server(
	authority,
	&committee,
	vec![],
	&[],
	network_config.buffer_size,
	)

sui/sui/src/sui_commands.rs

Lines 248 to 265 in 13c4e9f

	pub async fn make_server(
	authority: &AuthorityPrivateInfo,
	committee: &Committee,
	preload_modules: Vec<Vec<CompiledModule>>,
	preload_objects: &[Object],
	buffer_size: usize,
	) -> SuiResult<AuthorityServer> {
	let mut genesis_ctx = genesis::get_genesis_context();
	make_server_with_genesis_ctx(
	authority,
	committee,
	preload_modules,
	preload_objects,
	buffer_size,
	&mut genesis_ctx,
	)
	.await
	}

[patrickkuo]

just one comment on start_server function name

[gdanezis]

In PR #835 I point out (and fix) a bug in this logic: we need to check that the database is empty before calling again the logic to init it. You should be able to land this and I will rebase.