Avoid unnecessary allocation in Uncancelable trait.

[nmichael44]

Avoid unnecessary allocation in Uncancelable trait for idPoll. We can just allocate it once in object MonadCancel and apply the usual Id Monad trick (as in for example Resource.liftK).

[nmichael44]

This fails one of the tests with:

[error] cats-effect-kernel: Failed binary compatibility check against org.typelevel:cats-effect-kernel_3:3.4.10 (e:info.apiURL=https://typelevel.org/cats-effect/api/3.x/, e:info.versionScheme=early-semver)! Found 1 potential problems
[error]  * synthetic static method $init$(cats.effect.kernel.MonadCancel#Uncancelable)Unit in interface cats.effect.kernel.MonadCancel#Uncancelable does not have a correspondent in current version
[error]    filter with: ProblemFilters.exclude[DirectMissingMethodProblem]("cats.effect.kernel.MonadCancel#Uncancelable.$init$")
[error] java.lang.RuntimeException: Failed binary compatibility check against org.typelevel:cats-effect-kernel_3:3.4.10 (e:info.apiURL=https://typelevel.org/cats-effect/api/3.x/, e:info.versionScheme=early-semver)! Found 1 potential problems

This happens since there are no longer any fields in the trait and so the Scala compiler eliminates the $init$ method completely. So the new trait is missing a method, and that's why the test complains that the old and new trait definitions are incompatible.

Please advice what the right thing to do is in this case.

[durban]

I fear this is probably a valid complaint by mima. I suspect every existing implementer of MonadCancel#Uncancelable calls this $init$, so if it's missing, that's a problem. I'm not sure this minor optimization is worth it, if it causes a bincompat issue. (Although it is possible that there is some trick which would avoid the incompatibility...)

[nmichael44]

@durban Hi Daniel, I found a workaround to the mima problem. Please take a look and let me know when you get a chance.