Fix Issue 1004: effect order in filterAMissing

[j6carey]

The order of Applicative effects in filterAMissing was incorrect, causing the order of effects to differ from key order and be influenced by how the binary tree was balanced.

The fix is to arrange that effects arising from the key and value at an internal node come after those in its left branch instead of before. (Regardless of this fix such effects come before those effects arising from the right branch.)

This change also expands test coverage to detect a regression of this fix.

[j6carey]

Since I lack permission to initiate a continuous integration workflow or assign a reviewer, would someone be so kind as to trigger those steps?

[treeowl]

Sorry I missed this.

[treeowl]

Thanks!

[j6carey]

Thanks for the CI and merge!