scheduler: gate send_after_to runtime tests out of the loom build

These three RFC-015 tests call run() (the real runtime), so under
--cfg loom they construct loom atomics outside a loom::model block and
panic with "cannot access Loom execution state from outside a Loom
model". They are unit tests of runtime behavior, not state-machine
models. Gate the module #[cfg(all(test, not(loom)))], matching the
existing run_queue::tests precedent: they still run under normal
cargo test, and the loom build is now 28/28 clean.

src/scheduler.rs

@@ -726,7 +726,7 @@ pub fn run<F: FnOnce() + Send + 'static>(f: F) {
     crate::runtime::init(crate::runtime::Config::exact(1)).run(f);
 }
 
-#[cfg(test)]
+#[cfg(all(test, not(loom)))]
 mod send_after_to_tests {
     use super::*;
     use crate::channel::channel;