Adding a line that got lost in a recent merge.

1. Update to documentation concerning the running of until tests
   which code is compiled with fuzzing enabled.
2. Correction of text of fuzzing-related initialization exception.

1. The switch to enable fuzzing is now "--enable-fuzzing".
2. A check is now made to see that the CXX environment variable
   points to an AFL-enabled compiler when the switch is modified.

The convention is that all packets used in fuzzing have the file extension
.pkt.  This commit has renamed some packets to match that convention.

Add a doxygen file describing the fuzzing - how to build and
run it, and a brief description of the changes made to the code.

Instead of using a separate thread to read input from the fuzzer,
the input is now read in the main thread and transferred to the
interface on which Kea is expecting it to appear.

1. Convert to C++ class.
2. Extend framework to support DHCPv4 fuzzing.
3. Restrict size of data that can be accepted from AFL. (Kea will
   only accept up to about 64k or data, (set by the size of a UDP
   datagram).  However, AFL can send much larger data packets,
   which may cause problems in synchronization between the two
   threads used to implement fuzzing in Kea.

Proof of concept code to fuzz Kea-DHCPv6.

- added this support in hammer.py
- fixed checking sphinx in configure.ac on fedora 31