perf(devices): optimize memory allocations in virtio hot paths

Reduce heap allocations in performance-critical virtio device processing:

1. vsock_windows.rs:
   - Use fixed 4KB stack buffer instead of Vec in harvest_stream_reads
   - Eliminates heap allocation on every stream read iteration

2. net_windows.rs:
   - Use stack array for virtio-net header (was Vec)
   - Pre-allocate frame_data with 1500-byte capacity (typical MTU)
   - Remove unnecessary descriptor collection (iterate directly)
   - Remove unused DescriptorChain import

These changes reduce allocator pressure in the TX/RX hot paths without
changing behavior. All WHPX tests pass.

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

Author: hikejs

src/devices/src/virtio/net_windows.rs

@@ -20,7 +20,7 @@ use utils::eventfd::{EventFd, EFD_NONBLOCK};
 use vm_memory::{Bytes, GuestAddress, GuestMemoryMmap};
 
 use super::{
-    ActivateError, ActivateResult, DescriptorChain, DeviceState, InterruptTransport, Queue,
+    ActivateError, ActivateResult, DeviceState, InterruptTransport, Queue,
     VirtioDevice, TYPE_NET,
 };
 
@@ -183,12 +183,11 @@ impl Net {
         while let Some(head) = self.queues[TX_INDEX].pop(mem) {
             let index = head.index;
             let mut total_len: u32 = 0;
-            let mut hdr_bytes = vec![0u8; VIRTIO_NET_HDR_SIZE];
+            let mut hdr_bytes = [0u8; VIRTIO_NET_HDR_SIZE];
             let mut hdr_bytes_read: usize = 0;
-            let mut frame_data = Vec::new();
+            let mut frame_data = Vec::with_capacity(1500); // Pre-allocate for typical MTU
 
-            let descs: Vec<DescriptorChain<'_>> = head.into_iter().collect();
-            for desc in &descs {
+            for desc in head.into_iter() {
                 if desc.is_write_only() {
                     continue;
                 }

src/devices/src/virtio/vsock_windows.rs

@@ -31,6 +31,7 @@ const QUEUE_SIZE: u16 = 256;
 
 const VIRTIO_F_VERSION_1: u32 = 32;
 const VIRTIO_F_IN_ORDER: usize = 35;
+#[allow(dead_code)] // Reserved for future DGRAM implementation
 const VIRTIO_VSOCK_F_DGRAM: u32 = 3;
 const VSOCK_HOST_CID: u64 = 2;
 
@@ -547,8 +548,8 @@ impl Vsock {
 
                 // Read up to the available credit amount
                 let read_size = (available_credit as usize).min(4096);
-                let mut rx_buf = vec![0_u8; read_size];
-                match state.stream.read(&mut rx_buf) {
+                let mut rx_buf = [0u8; 4096];
+                match state.stream.read(&mut rx_buf[..read_size]) {
                     Ok(n) if n > 0 => {
                         // Update tx_cnt to track bytes sent to peer
                         state.tx_cnt = state.tx_cnt.saturating_add(n as u32);