Improve netlink interface code and packet validation

The current logic for using `netlink(7)` sockets and parsing packets does not do proper validation on the length of the received packet. The parsing code uses `libnl` to parse the packets, but calls into `nlmsg_parse()` directly without first validating the packet with `nlmsg_ok()`. Because `recv(2)` defaults to silently truncating netlink packets that are larger than the received buffer, we may receive a packet with too many attributes that extends beyond the buffer we passed in. When we begin to parse the packet, we may end up doing an out-of-bounds memory access due to this. The `nlmsg_parse()` API function does not take in the whole packet length as a parameter and assumes the packet was already validated with `nlmsg_ok()` or similar logic. This bug occurs at least twice in `lldp_util.c` and also in `qbg/vdpnl.c`. Part of this is currently being fixed in PR #122. Further use of netlink should be investigated as well.

The documentation for `libnl` at https://www.infradead.org/~tgr/libnl/doc/core.html might be worth a read and perhaps we should be using more of the library to handle some of the logic we are already doing. For example, some of the logic proposed to allocate buffers is provided by `nl_recv()`.