ext-http: SIGABRT 'there is no reactor running' (server/server.rs:913) — node:http server + ws WebSocketServer + fetch

[proggeramlug]

Env: perry 0.5.1512 built from main@6173ff9f6, macOS arm64. MB24 Phase-0 spike.

Repro:

import { createServer } from "node:http";
import { WebSocketServer } from "ws";
const httpServer = createServer((_req, res) => { res.writeHead(200); res.end("http-ok"); });
const wss = new WebSocketServer({ server: httpServer });
await new Promise<void>((r) => httpServer.listen(8137, r));
console.log("listening");                       // reached
const res = await fetch("http://127.0.0.1:8137/"); // crash while serving this

Actual: hard abort (exit 134):

thread '<unnamed>' (49791956) panicked at crates/perry-ext-http/src/server/server.rs:913:38:
there is no reactor running, must be called from the context of a Tokio 1.x runtime

Expected: plain HTTP requests served alongside a ws upgrade handler on the same server (the standard single-port WS + REST layout).

Note: upstream main has 4 newer ext-http commits (#8615/#8608/#8632/#8641) but they target the client/Agent side; the panic site is in the server path. Full test: mobilitybroker24/monorepo spike/tests/07-ws.ts.

[proggeramlug]

Closed by #8769, landed on main via #8778 (squash 1f5c3bb80).

That PR was landed through a batch wrapper PR rather than merged directly, so GitHub's close keyword never fired at merge time. Verified before closing: changelog.d/8769-http-ws-reactor-scheduling.md is present on main, and the accept loops now schedule through the reactor-owned async bridge.