Fix gethostbyname(3) data race

[meme]

Under certain conditions, such as a network failure, socket_connect_timeout will race to call gethostbyname(3) and end up dereferencing NULL:

if ((hp = gethostbyname(addr)) == NULL) {
    usbfluxd_log(LL_ERROR, "%s: unknown host '%s'", __func__, addr);
    return -1;
}

// Passes, data is fine
if (!hp->h_addr) {
    usbfluxd_log(LL_ERROR, "%s: gethostbyname returned NULL address!", __func__);
    return -1;
}

// ...
// Another thread calls gethostbyname, hp is now pointing to another instance (data race)
// ...

// BOOM
saddr.sin_addr.s_addr = *(uint32_t *) hp->h_addr;

The man page makes this apparent:

   The  functions  gethostbyname()  and gethostbyaddr() may return pointers to static data, which may be overwritten by
   later calls.  Copying the struct hostent does not suffice, since it contains pointers; a deep copy is required.

ASan:

==28115==ERROR: AddressSanitizer: SEGV on unknown address 0x000000000000 (pc 0x561c82fb46e6 bp 0x7f8520a20ca0 sp 0x7f8520a20ae0 T357)
==28115==The signal is caused by a READ memory access.
==28115==Hint: address points to the zero page.
    #0 0x561c82fb46e6 in socket_connect_timeout /usbfluxd/usbfluxd/socket.c:153
    #1 0x561c82fba741 in check_remote_func /usbfluxd/usbfluxd/usbmux_remote.c:949
    #2 0x7f85da806258 in start_thread (/usr/lib/libpthread.so.0+0x9258)
    #3 0x7f85da5c75e2 in __GI___clone (/usr/lib/libc.so.6+0xfe5e2)

AddressSanitizer can not provide additional info.
SUMMARY: AddressSanitizer: SEGV /usbfluxd/usbfluxd/socket.c:153 in socket_connect_timeout
Thread T357 created by T0 here:
    #0 0x7f85da889fa7 in __interceptor_pthread_create /build/gcc/src/gcc/libsanitizer/asan/asan_interceptors.cpp:216
    #1 0x561c82fbaa4b in usbmux_remote_get_fds /usbfluxd/usbfluxd/usbmux_remote.c:967
    #2 0x561c82fc0ef2 in main_loop /usbfluxd/usbfluxd/main.c:129
    #3 0x561c82fc2870 in main /usbfluxd/usbfluxd/main.c:477
    #4 0x7f85da4f0b24 in __libc_start_main (/usr/lib/libc.so.6+0x27b24)

[strazzere]

LGTM, much appreciated.