Actually use scheduler in reverse rules

[lkdvos]

This is a small change that ensures that we are actually using the correct scheduler in the reverse rules. I don't really understand why or how this is happening, but I noticed that the current way the rrule was implemented seems to intercept the callstack before the default keyword gets filled in, therefore always using the default scheduler, even if set to :serial.

[codecov]

Codecov Report

✅ All modified and coverable lines are covered by tests.

Files with missing lines	Coverage Δ
src/utility/diffable_threads.jl	100.00% <100.00%> (ø)

🚀 New features to boost your workflow:

• ❄️ Test Analytics: Detect flaky tests, report on failures, and find test suite problems.

[pbrehmer]

Thanks for fixing that! I also don't get why the kwargs of dtmap are not being passed onto its rrule. However, I vaguely remember having problems with this before and IIRC I had to explicitly pass the kwargs to the forward function call, i.e. the default kwargs wouldn't be used somehow.

Anyways, now that the schedulers of the forward and backward pass are essentially independent, what would you think of having two separate forward and backward schedulers? I bring this up since I was worried that multi threading the backward pass might be quite inefficient in some cases and then it would be useful to set just the backwards scheduler to :serial. In any case that would be another PR, so this is ready to go for me.

[lkdvos]

I’m definitely okay with having both separately, I’ll see what I can do