fix allocations introduced by #212

[Krastanov]

In #212 we introduced compactification for ClassicalXOR which seems to cause more allocations (see the benchmark comments in that issue). Fix those and add explicit tests confirming we have low allocations (not simply benchmarks) for the functionality in question