Use lazy sense calculator

[esseivaju]

The lazy sense calculators require potentially recalculating the same sense multiple times. However, they do not use global memory so they have the potential to be faster, especially once we move to infix evaluation and short-circuiting.

I am splitting this from #1565 to test whether we see any performance improvement without short-circuiting.

[github-actions]

Test summary

 4 433 files   6 859 suites   15m 20s ⏱️
 1 683 tests  1 677 ✅  6 💤 0 ❌
24 092 runs  24 008 ✅ 84 💤 0 ❌

Results for commit 8308a19.

♻️ This comment has been updated with latest results.

[esseivaju]

Not sure why only that one unit test would fail 🤔

[esseivaju]

Timings are essentially the same as in #1569 (comment), complex_intersect doesn't make a difference.

[esseivaju]

@sethrj Do you want to merge this, or should we do everything as part of the infix eval PR?

[sethrj]

Thanks @esseivaju !