fix(ci): address PR #35 review comments

- Document non-blocking receive contract in perf_bench.rs docstring,
  confirming no TCP deadlock race (Hoshe #1, critical)
- Make shadowcast parser order-independent — flush on new config header
  instead of requiring Recursive after Symmetric (Hoshe #2)
- Fix p95 calculation: use floor(0.95*(N-1)) nearest-rank instead of
  ceil(0.95*N)-1 which was off-by-one at N=50 (Hoshe #4)
- Error on --compare when no baseline file exists (Hoshe #5)
- Add D-031 10tps assumption comment to TICK_BUDGET_US (Tyre #1)
- Strengthen snapshot assertion: require warmup + half measurement
  window instead of warmup + 1 (Tyre #2)
- Regenerate baseline with corrected p95 (356µs, was 526µs)

Hoshe #3 (.PHONY) was already addressed — perf-baseline is in the
.PHONY declaration on Makefile line 10.

Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>

server/tests/perf_bench.rs

@@ -55,6 +55,12 @@ fn read_rss_kb() -> Option<u64> {
 /// Boots the server with production content, runs WARMUP_TICKS to stabilize,
 /// then measures MEASURE_TICKS of app.update() wall-clock time. Reports entity
 /// counts from observer snapshots and process RSS.
+///
+/// Protocol contract: BridgePlugin uses non-blocking receive (WouldBlock →
+/// empty input vec), so the server always advances even if the client hasn't
+/// sent input yet. The server sends a snapshot each tick; the client blocks
+/// on read until one arrives, then responds with (empty) input. No deadlock
+/// possible — see TcpBridge::receive_inputs and send_snapshot in bridge/tcp.rs.
 #[test]
 #[ignore]
 fn perf_tick_timing() {
@@ -145,12 +151,16 @@ fn perf_tick_timing() {
         }
     }
 
-    // Must have received enough data past warmup for meaningful results
+    // Must have received enough measured snapshots for meaningful results.
+    // Require all warmup ticks plus at least half the measurement window.
+    let min_snapshots = WARMUP_TICKS + MEASURE_TICKS / 2;
     assert!(
-        entity_counts.len() > WARMUP_TICKS,
-        "Only received {} snapshots, need at least {} (warmup) + 1",
+        entity_counts.len() >= min_snapshots,
+        "Only received {} snapshots, need at least {} ({} warmup + {} measured)",
         entity_counts.len(),
-        WARMUP_TICKS
+        min_snapshots,
+        WARMUP_TICKS,
+        MEASURE_TICKS / 2
     );
 
     drop(reader);
@@ -174,7 +184,8 @@ fn perf_tick_timing() {
 
     let mut sorted = measured_us.clone();
     sorted.sort();
-    let p95_idx = ((sorted.len() as f64 * 0.95).ceil() as usize).saturating_sub(1);
+    // Nearest-rank p95: index = floor(0.95 * (N-1)) for 0-based indexing.
+    let p95_idx = ((sorted.len() - 1) as f64 * 0.95).floor() as usize;
     let p95 = sorted[p95_idx.min(sorted.len() - 1)];
 
     let entity_counts_measured: Vec<usize> =