fix: correct BPE merge annotations in test_wikipedia_example

[coding-shalabh]

Summary

• Fixed incorrect BPE merge sequence annotations in test_wikipedia_example docstring
• The docstring claimed X=ZY, Y=ab, Z=aa but the actual BPE algorithm produces Z=aa, Y=Za (aaa), X=Yb (aaab)
• Added explicit merge sequence with occurrence counts for clarity

Details

The original annotation described the merges as:

X=ZY
Y=ab
Z=aa

But running BPE on "aaabdaaabac" for 3 merges actually produces:

1. (97, 97) -> 256 (Z=aa), most frequent pair (4 occurrences)
2. (256, 97) -> 257 (Y=Za=aaa), most frequent pair (2 occurrences)
3. (257, 98) -> 258 (X=Yb=aaab), most frequent pair (2 occurrences)

This can be verified with:

tokenizer = BasicTokenizer()
tokenizer.train("aaabdaaabac", 256 + 3, verbose=True)
# merge 1/3: (97, 97) -> 256 (b'aa') had 4 occurrences
# merge 2/3: (256, 97) -> 257 (b'aaa') had 2 occurrences
# merge 3/3: (257, 98) -> 258 (b'aaab') had 2 occurrences

The test assertions were already correct ([258, 100, 258, 97, 99]), only the docstring was wrong.

Fixes #93

Test plan

• pytest tests/test_tokenizer.py::test_wikipedia_example — both BasicTokenizer and RegexTokenizer pass