feat(runtime): add workload scheduler micro-benchmark

Add workload-scheduler-micro-bench binary comparing poll share, E2E
wall latency, active exec time, and the new catio admission wait
under synthetic dual-backlog loads. 2:8 vs 1:1 runs confirm
admission wait and wall latency reflect the weights while poll share
does not (0.500/0.500 in both), so admission wait is the usable
fairness signal.

Signed-off-by: discord9 <55937128+discord9@users.noreply.github.com>

src/common/runtime/Cargo.toml

@@ -7,10 +7,21 @@ license.workspace = true
 [lib]
 path = "src/lib.rs"
 
+[features]
+default = []
+# Gate for optional benchmark/test-only binaries; empty on purpose.
+test = []
+
 [[bin]]
 name = "common-runtime-bin"
 path = "src/bin.rs"
 
+[[bench]]
+name = "workload-scheduler-micro-bench"
+path = "benches/workload_scheduler.rs"
+harness = false
+required-features = ["test"]
+
 [lints]
 workspace = true