tests: transformed tests/sha256 into a unittest

[phiros]

Partial fix for issue #1131.

tests/sha256 -> unittest/tests-crypto/tests-crypto-sha256.c

[miri64]

Would it make sense to devide those up into several tests?

Also: there is the macro TEST_ASSERT_EQUAL_STRING. With it the output in case of error looks a little bit more helpful.

[phiros]

Whenever I compare strings in C I prefer to use strncmp. Maybe thats just nit-picking but I think it
makes sense (especially in the context of a test). The TEST_ASSERT_EQUAL_STRING macro is essentially a strcmp.

[miri64]

I know this is just a tranformation of tests/tests-crypto/, but I would the unittests to be less bulky and testing one API function at a time. BTW: What's with corner cases?

[phiros]

For now I'd really prefer to get this PR merged so that I can ask the authors of this code the crypto module ( @mehlis and @Kijewski ) to improve this unit test (they know the implementation far better than I do; it would be easier for them to identify those corner cases).
Also, this PR would ensure that the test is at least executed. Currently, this is not the case.

BTW: As previously mentioned, the overall test coverage of crypto is quite poor (all 5 cipher implementations are untested). I think it makes sense to open a separate issue for this.

[LudwigKnuepfer]

In general a fresh mind might be better at finding corner cases..

[miri64]

I'm with @LudwigOrtmann. It's better if the test author is someone else than the author of the tested module, but since our manpower does not really allow for that (yet), I see where you coming from. Plus, this is not topical to this PR. Let's merge it: ACK (tested for native, samr21-xpro and msba2)