[WIP] Experiment compacting counttables into a nodetable given a max abundance criterion

[standage]

Scenario: I want to query some big ol' counttables, and my k-mers of interest are those that are abundance <= X. Rather than keeping several of these big counttables in memory at once, I'd like to be able to initialize and update a nodetable (with the same number of buckets) to indicate which k-mers satisfy the given criterion. That way I only have to keep one counttable in memory at a time.

This PR introduces two new methods for nodetables: compose_init, and compose_update. These methods do bit flipping directly on the byte array.

The build is currently failing. I'm guessing there's an issue at the C++ level, and maybe at the Cython level as well. This could really benefit from a preliminary review from @luizirber or @camillescott.

---

• Is any new functionality in tested? (This can be checked with
  make clean diff-cover or the CodeCov report that is automatically
  generated following a successful CI build.)
• Was a spellchecker run on the source code and documentation after
  changes were made?
• Have any substantial changes been documented in CHANGELOG.md? See
  keepachangelog for more details.