fix: Forward duration flag to spawned server in blocking runner

The blocking benchmark runner always passed --msg-count (returning 0
in duration mode) to the spawned server process, never forwarding -d.
This caused server/client config divergence in duration-based blocking
benchmarks — the server would receive --msg-count 0 while the client
ran for a specific duration.

Now mirrors the async runner's logic:
- If config.duration is set, pass -d <seconds>
- Otherwise if config.msg_count is set, pass --msg-count <count>
- Otherwise pass neither (server uses defaults)

Fixes #121

AI-assisted-by: Claude Opus 4.6
Co-authored-by: Cursor <cursoragent@cursor.com>

Author: mcurrier2

src/benchmark_blocking.rs

@@ -439,10 +439,16 @@ impl BlockingBenchmarkRunner {
         cmd.arg("-m")
             .arg(self.mechanism.to_possible_value().unwrap().get_name());
 
-        // Add message size and count
+        // Add message size
         cmd.arg("--message-size")
             .arg(self.config.message_size.to_string());
-        cmd.arg("--msg-count").arg(self.get_msg_count().to_string());
+
+        // Forward duration or msg-count (duration takes precedence)
+        if let Some(duration) = self.config.duration {
+            cmd.arg("-d").arg(format!("{}s", duration.as_secs_f64()));
+        } else if let Some(count) = self.config.msg_count {
+            cmd.arg("--msg-count").arg(count.to_string());
+        }
 
         // Add transport-specific identifiers
         if !transport_config.socket_path.is_empty() {