Fix endianness issues in sys::unix::in_addr_convertion test

Previously, the "expected" values were implicitly encoded in
little-endian byte order with bitwise operators to match the
native (little-endian) byte order of u32s on most architectures.
This test obviously breaks when the byte order is big-endian.

Instead, provide the expected value as an array of numbers
converted into a network-endian u32. This has the added benefit
that it's much clearer to read, and the numbers in the array
also match the ones that are used in the Ipv4Addr constructor, and
no longer have to be constructed manually with bitwise operators.

src/sys/unix.rs

@@ -1449,13 +1449,13 @@ fn in_addr_convertion() {
     let raw = to_in_addr(&ip);
     // NOTE: `in_addr` is packed on NetBSD and it's unsafe to borrow.
     let a = raw.s_addr;
-    assert_eq!(a, 127 | 1 << 24);
+    assert_eq!(a, u32::from_ne_bytes([127, 0, 0, 1]));
     assert_eq!(from_in_addr(raw), ip);
 
     let ip = Ipv4Addr::new(127, 34, 4, 12);
     let raw = to_in_addr(&ip);
     let a = raw.s_addr;
-    assert_eq!(a, 127 << 0 | 34 << 8 | 4 << 16 | 12 << 24);
+    assert_eq!(a, u32::from_ne_bytes([127, 34, 4, 12]));
     assert_eq!(from_in_addr(raw), ip);
 }