glib-aux: choose favorable reallocation sequence in nm_utils_fd_read()

Ensuring to have 1000 bytes available is not a good start.
Because, if we start with an empty buffer (allocated fewer
than 1000 bytes), on first call we will reserve 1000 bytes.
If we then just read one byte, on the next read we would again
try to reserve 1000 bytes, and already require a reallocation.

In this case it is better to reserve 488+1 or 1000+1 bytes.

src/libnm-glib-aux/nm-io-utils.c

@@ -482,13 +482,17 @@ nm_utils_fd_read(int fd, NMStrBuf *out_string)
     g_return_val_if_fail(fd >= 0, -1);
     g_return_val_if_fail(out_string, -1);
 
-    /* If the buffer size is 0, we allocate NM_UTILS_GET_NEXT_REALLOC_SIZE_1000 (1000 bytes)
-     * the first time. Afterwards, the buffer grows exponentially.
+    /* Reserve at least 488+1 bytes of buffer size. That is probably a suitable
+     * compromise between not wasting too much buffer space and not reading too much.
      *
-     * Note that with @buf_available, we always would read as much buffer as we actually
-     * have reserved. */
-    nm_str_buf_maybe_expand(out_string, NM_UTILS_GET_NEXT_REALLOC_SIZE_1000, FALSE);
+     * Note that when we start with an empty buffer, the first allocation of
+     * 488+1 bytes will actually allocate 1000 bytes. So if we were to receive
+     * one byte at a time, we don't need a reallocation for the first 1000-(488+1)
+     * bytes. Afterwards grows the buffer exponentially.
+     */
+    nm_str_buf_maybe_expand(out_string, NM_UTILS_GET_NEXT_REALLOC_SIZE_488 + 1, FALSE);
 
+    /* We always use all the available buffer size. */
     buf_available = out_string->allocated - out_string->len;
 
     n_read = read(fd, &((nm_str_buf_get_str_unsafe(out_string))[out_string->len]), buf_available);