event loop: run the timers phase before the entry point's queued work - #33509

🟡 The worker path's guard at web_worker.rs:1144 checks `vm.unhandled_error_counter == 0` before `drain_expired_timers()`, but `run_entry_point_body` only checks `status_ptr == Rejected` — so identical `Promise.reject(...); setTimeout(..., 1); /*busy-wait*/` code now fires the timer under `bun run`/`bun test` but not in a worker (both matched pre-PR: neither fired). After 57ff06aa switched to `self.tick()`, that tick's trailing `handle_rejected_promises()` has already bumped the counter by the ti

The body of `__bun_drain_expired_timers` is byte-identical to the pre-existing `drain_timers_export` at `src/runtime/timer/Timer.rs:503-511` (both: `timer_all()` → null-check → `(*all).drain_timers(vm.cast::<()>())`). Since both live in `bun_runtime`, the new `#[no_mangle]` shim could be a one-line `crate::timer::timer::drain_timers_export(vm)` delegation instead of duplicating the null-guard and cast — otherwise a future change to either has to be made in two places.

The doc comment here says "(see the Known-exclusion note on the PR)", and node-timers.test.ts:255 says "see the PR's Known-exclusion note" — after merge, "the PR" is unresolvable from source without git-blame. Suggest dropping both parentheticals (the sentences already state the exclusion self-contained, as the sibling web_worker.rs comment in this PR does), or use the numbered form `PR #33509` matching existing precedent in `src/`.