CI-unixish.yml: use LTO in the callgrind step

[firewave]

[firewave]

This is to make sure that we (well, me) do not try to mess with code which can actually be optimized by the compiler.

From my local system with CXX=g++ VALGRIND_TOOL=callgrind SIMPLECPP_PATH=simplecpp-1.5.1 ./selfcheck.sh:

g++ -O2 - 66,196,507 / 1,217,069,685 g++ -O2 -flto - 64,854,450 / 1,185,122,059 g++ -O3 - 65,023,794 / 1,198,427,777 g++ -O3 -flto - 65,153,503 / 1,190,654,841 clang++ -O2 - 67,422,122 / 1,227,600,090 clang++ -O2 -flto - 66,823,868 / 1,213,656,996 clang++ -O3 - 65,293,400 / 1,177,431,242 clang++ -O3 -flto - 64,801,161 / 1,164,063,469

The differences could well just be inaccuracies within callgrind but I have not looked into it yet.