Optimize copy method from O(N log N) to O(N)

[HossamSaberr]

Fixes #37

This PR optimizes the copy functionality of CTAVLTree by replacing the O(N log N) element-by-element re-insertion with a O(N) structural deep copy.

Changes:

1. Removed the custom copy method in CTAVLTree that relied on inOrderDo:, allowing the system to use the standard shallowCopy -> postCopy.
2. Implemented postCopy in CTAVLTree to initiate the root duplication.
3. Implemented a recursive copy in CTAVLNode that directly duplicates left and right children while manually injecting the height instance variable, completely bypassing add: balancing math and rotations.
4. Added an instant copy return for CTAVLNilNode.
5. Added test cases (testCopyEmptyTree, testCopySingleNode, testCopyStructureAndIsolation, testCopyLargeTree) to prove structural integrity, exact shape matching, and memory isolation.

Performance Proof:
Benchmarking the old implementation vs. the new O(N) implementation shows constant factor reduction and linear scaling:

N (Elements)	Old copy O(N log N)	New copy O(N)	Improvement
10,000	9 ms	4 ms	~2.2x faster
20,000	22 ms	9 ms	~2.4x faster
40,000	40 ms	33 ms	~1.2x faster
80,000	190 ms	63 ms	~3.0x faster
160,000	339 ms	224 ms	~1.5x faster

Benchmark Script & Visual Proof

You can reproduce these results using the following Playground script:

| sizes results |
sizes := #(10000 20000 40000 80000 160000).
results := OrderedCollection new.

sizes do: [ :n |
    | tree data copyTime |
    tree := CTAVLTree new.
    data := (1 to: n) asArray shuffled.
    tree addAll: data.
    
    copyTime := Time millisecondsToRun: [ tree copy ].
    
    results add: ('N=', n asString, ' took ', copyTime asString, ' ms').
].

results inspect.

Before (O(N log N)):

After (O(N)):

[HossamSaberr]

Done, I've added testCopyLeafNode to ensure the individual nodes copy correctly. I also took the opportunity to fix the CI failure by refactoring CTAVLNode >> copy into postCopy, which is much cleaner anyway. The CI should be Green now