Draft: Reproduce #1242 in shadow ci

• Increase simulation time from 30m to 5h.
• Increase shadow's heartbeat-interval and host-heartbeat-interval to reduce bookkeeping overhead.
• Remove unrelated hosts from the experiment to reduce overhead.
• Add additional articlient-onion-artiserver-* clients that try to use the HS at later times.

The simulation normally uses two builds for arti - a "vanilla" build (normally built in ci job "rust-latest") and an "extra" build with extra features (normally built in ci job "rust-latest-arti-extra-features").

When reproducing locally I usually just use the "extra" build for both:

$ cargo build --verbose --target x86_64-unknown-linux-gnu -p arti -p tor-circmgr --bin arti --features full,experimental-api,arti-client/keymgr,tor-circmgr/ntor_v3,onion-service-service
$ cp target/x86_64-unknown-linux-gnu/debug/arti target/x86_64-unknown-linux-gnu/debug/arti-extra
$ cd tests/shadow
$ ./run.sh

tests/shadow/run.sh

@@ -14,13 +14,15 @@ chmod 700 shadow.data.template/hosts/fileserver-onion/hs
 chmod 700 shadow.data.template/hosts/fileserver-onion-auth/hs
 
 # Run the simulation
+#  --use-worker-spinning=false \
 shadow \
-  --log-level=debug \
+  --heartbeat-interval=30m \
+  --host-heartbeat-interval=30m \
+  --log-level=info \
   --strace-logging-mode=standard \
   --template-directory=./shadow.data.template \
   --progress=true \
   --use-memory-manager=false \
-  --use-worker-spinning=false \
   shadow.yaml \
   > shadow.log