Do we need structs in plan::tracing and ProcessedgesWork to be public?

[qinsoon]

I noticed these API changes reported in #897. The changed items are exposed as public. But it is probably unnecessary to make them public. We should consider making them private.

Changed items in the public API
===============================
-pub fn mmtk::plan::ObjectsClosure<'a, E>::new(worker: &'a mut mmtk::scheduler::GCWorker<<E as mmtk::scheduler::ProcessEdgesWork>::VM>) -> Self
+pub fn mmtk::plan::ObjectsClosure<'a, E>::new(worker: &'a mut mmtk::scheduler::GCWorker<<E as mmtk::scheduler::ProcessEdgesWork>::VM>, bucket: mmtk::scheduler::WorkBucketStage) -> Self
...
-pub fn mmtk::scheduler::ProcessEdgesWork::new(edges: alloc::vec::Vec<<<Self as mmtk::scheduler::ProcessEdgesWork>::VM as mmtk::vm::VMBinding>::VMEdge>, roots: bool, mmtk: &'static mmtk::MMTK<Self::VM>) -> Self
+pub fn mmtk::scheduler::ProcessEdgesWork::new(edges: alloc::vec::Vec<<<Self as mmtk::scheduler::ProcessEdgesWork>::VM as mmtk::vm::VMBinding>::VMEdge>, roots: bool, mmtk: &'static mmtk::MMTK<Self::VM>, bucket: mmtk::scheduler::WorkBucketStage) -> Self

[wks]

It's likely a bug. It should be enough to keep them pub(crate).

VM bindings don't need to depend on the ProcessEdgesWork trait in order to work. See #599