compatibility-v2

- the lognormalizer tool now defaults to json output format
- API changes
  * ln_loadSample() does no longer exist
    loading of samples at runtime is no longer possible. This is due to
    the fact that the parse DAG is compiled into the most efficient form
    and thereafter read-only. If urgently needed, we could change the code
    so that the unoptimized form is kept (memory intense) and a re-compile
    happens after each addition (time intense).
- the plain old iptables parser is no longer supported. Sorry for that,
  but keeping it would have caused violation of layers. It can be
  replaced by the new one
- the rulebase format for v2 has changed to support the enhancements

  It is mostly compatible with v1, which means it understands mosts of
  it's constructs. However, some are NOT understood:
	- the "tokenized" parser is no longer supported. Use the more
	  capable "repeat" parser instead.
	- the "recursive" and "descent" parsers are no longer supported. Use
	  user-defined data types instead.
	- the suffixed parser and friends are no longer supported. A
	  replacement is currently being developed.
	- the regexp parser is no longer supported and needs to be
	  replaced by other liblognorm features
	  for details on this decision see:
	  https://github.com/rsyslog/liblognorm/issues/143

  To signify that a rule base file contains v2 format, it must contain
  the line

  version=2

  as the very first line of the file, in exactly this format (no comments,
  no whitespace, nothing else in between the words or after them). If that
  line is missing or not given **exactly** as above, the old v1 engine is
  used, with all of its restrictions. If it is present, the v2 engine is
  used.