regression: CI broken because of recent not unary change

[timotheecour]

/cc @Araq
https://travis-ci.org/nim-lang/Nim/jobs/452152357

all PR's are now broken following 8a357c2
although I really liked that we could now do doAssert not foo ...

FAIL: tinvalid_construction.nim C
Test "tests/constructors/tinvalid_construction.nim" in category "constructors"
Failure: reNimcCrash
Expected:

there's quite a few recent instances like that where CI broke after a manual merge that would've been caught by CI, had it been run; I know it's a pain to wait for CI, but hopefully #9581 will enable making it a lot faster to run