Quick fix for the second half of the assertion bug

[joka921]

• same fixing strategy also for the function that gets the following block.

• Also fixed problems in case the scanNonFunctionalRelation etc. functions get an empty input.

• the whole bloated scanPOS, scanPSO, ... functions in the Index class clearly need refactoring, but not tonight.

[niklas88]

Besides the scanXXX() methods I think the two functions we touched for this bug are really hard to grasp too. But I think this fix is minimally invasive and I've tested that this indeed fixes #206.

That said I still don't understand the logic around the it == _blocks.end() case. In that case we decrement it, then don't increment it again and then set after to _startRhs. That however means we get the exact same result as for getBlockStartAndNofBytesForLhs() (as we are skipping the increment which is the only difference). Does that make sense is the last block its own follow block?

EDIT: Looking at the docs for std::lower_bound() I was confused for a bit how we could end up with it == _blocks.end() I understand this only happens for an Id larger than any in the _blocks. Shouldn't that also mean we have an empty result?

EDIT2: Nevermind it's in the function comment for getFollowBlockForLhs()

[niklas88]

As you said these functions as well as scanXXX() are quite confusing and duplicate code. But this is a task for another day and this fix is minimally invasive and fixes #206 so let's merge this now.