Fixes #1 (closed).

@anoe The bug was a bit subtle and easy to miss: in the call to igraphCommunityInfomap we were passing as the last argument for *codelen the nullPtr. However, by looking at the C implementation of this function, we can see how we need a valid pointer as the function would write at that address the result of the calculation (igraph_real_t) shortestCodeLength / log(2.0).

In terms of "how do I fix this next time", alas I don't have a silver bullet here. First of all, the first step is to make sure we are indeed getting a segfault. Unfortunately neither stack nor cabal would be very helpful in this regards when running the tests via stack test or cabal test. However, more informative logs could be acquired by running directly the test executable (essentially bypassing stack or cabal). I'm a cabal user (but for stack it should be equivalent), so I was able to run something like dist-newstyle/build/x86_64-osx/ghc-8.10.7/haskell-igraph-0.8.5/t/test/build/test (the path would be slightly different on your machine) to be greeted with a segfault message, whereas by running this via cabal v2-test was simply giving me a FAIL test without an indication of why that was (perhaps the segfault is emitted to stderr, albeit I didn't check that was the case).

In theory it should also be possible to throw gdb at this, but at least on my Mac gdb is a bit all over the place and not really working well, but luckily a code review was enough here to spot the problem 😉