Make reduction accumulators not inherit predicates

[inducer]

Closes #231. Closes inducer/pytential#57.

[inducer]

The hope is that this will address inducer/pytential#57. The first round of CI passed (including the pytential downstream test). I'll rerun to gather more data.

[kaushikcfd]

As long as the assignment instruction is surrounded by the predicate this should be fine. LGTM, thank you!

[inducer]

@kaushikcfd Thanks for taking a look!

🎉 Passed a second time, too. I'll do another run and click auto-merge for that, so that if it passes, it goes in.

[inducer]

Hmm, well, technically, these tests don't run test the Pytential example that was failing... so I created a dummy pytential PR: inducer/pytential#64

I guess I'd like to get a few runs of that in to see how it behaves. If it passes... maybe three...? of those, I'd call it good to go.

[inducer]

inducer/pytential#64 passed three times. I'll call that good enough for me.