[FEA] Explore using Cython language features that allow minimizing overhead and simplifying code

**Is your feature request related to a problem? Please describe.**
Cython offers various small bits of functionality that reduce overhead in various scenarios. We should test some of these out.

**Describe the solution you'd like**
Features to test:
- [ ] Apply `cython.final` decorator to classes. This will reduce the overhead of `cpdef` methods by removing the need to look up child classes.
- [ ] Use `cython.freelist` for objects that may be created and freed often. Good examples would be lightweight objects like `pylibcudf.aggregation.Aggregation`s
- [ ] Add `not None` to parameters that must be provided so that they can be checked cheaply at runtime.