This reverts commit db606f7.

Conflicts:
	tck/src/main/resources/spec.md

Includes:
* Initial tests for spec rules for Publisher / Subscriber
* The Processor specs delegate to all pub / sub specs
* Additional Processor rules are REMAINS from the OLD TCK,
  and have not yet been pruned / cleaned up.
* Added travis to build; there is only one test which validates that all
  pub / sub rules are properly delegated to from the processor test.
* Tests now fail in isolation, previous TCK could cascade one test
  failure to multiple tests failing

Please review and share feedback, it's still a bit early and needs more polish.
The classes I would ask you to focus on are PublisherVerification and
SubscriberVerification.

* Make spec rules linkable (#2.13) and use links in TCK
* Verified 1-element-publishers are testable with TCK (skips unsupported tests)
+ moved 3.2 from subscriber tests to publisher tests
+ moved 3.4 to publisher
+ moved 3.6 spec to Publisher
+ check rule 3.3, assuming publisher breaks the reccursion
+ Splitting subscriber whitebox / blackbox tests
+ both got verifications of rule 2.3
+ whitebox is the previous impl
+ blackbox is tests which work without any additional probes
  it mostly wrapps the subscriber in order to check things
+ better 2.10 verification in blackbox
+ test was testing onComplete instead of onError like it should
+ better blackbox tests for 2.9
+ more java-doc style java-docs (added {@link}s)
+ minor cleanups and refactors to make BlackboxProxy more understandable
+ added missing delegation of boundedDepthOfOnNextAndRequestRecursion
+ improved error message when expecting error message

…reams

+ checking 3.9 in PublisherVerification
+ better error messages on expectThrowingOf
+ fixed 2.5 tests – they were checking the wrong things
+ checking 1.13 must be optional, because some publishers dont support fan-out
+ inverse condition in 3.13 to avoid NPE, and report a nice error
+ checking 1.7 should give more demand, then pull based publishers will
  notice that their source has been drained and complete the downstream
+ checking 1.13 should give more demand, as if the publisher is
  pull-based, it would not notice (and has 0 buffer space), it would not
  notice that it's source has been drained, and would not complete the
  stream.

+ 3.9 check; fixes typo introduced during 0 semantics change
+ fixed typo in error string
+ reworked 2.12 test

Rationale:
It was part of the previous TCK, but I find it hard to support for
different kinds of Publishers we might have. One example being "hot"
Publishers. One might argue that the TCK must then silence this test
as it can not "buffer" the required amount of elements. Also, this test
is not directly testing TCK rules directly (but involves the in-order
requirements and buffering implicitly)...

if your impl is not able to support multiple subscribers