Exercise 1: Extending the Insert-Timing Measurement to 3,000 Rows — Possible Solution
====================================================================

THE TEST
------------------------------
    t500 = measure_full_rewrite_inserts(500)
    t3000 = measure_full_rewrite_inserts(3000)
    ratio = t3000 / t500

RESULT
------------------------------
    500 inserts:  0.165s
    3000 inserts: 1.877s
    ratio: 11.37x

3,000 inserts is 6x the data of 500 inserts. A linear relationship
would predict the time also grows by exactly 6x. The real, measured
ratio is 11.37x -- nearly double what linear scaling would predict.

WHY THE RATIO IS WORSE THAN 6x, NOT JUST DIFFERENT
------------------------------
naive_insert_full_rewrite's own per-insert cost isn't constant -- it
grows with however many rows are ALREADY stored, since every insert
re-serializes the entire accumulated list. Insert number k has to
write out k rows, not 1. Summing that cost across N total inserts
(1 + 2 + 3 + ... + N) gives a total proportional to N(N+1)/2 -- a
quadratic function of N, not a linear one. For N=500, that sum is
about 125,250; for N=3000, it's about 4,501,500 -- a ratio of almost
exactly 36x (6 squared), not 6x. The measured 11.37x is smaller than
that theoretical 36x (real-world I/O has fixed per-write overhead that
doesn't grow with data size, plus OS/filesystem caching effects that
blunt the pure math), but it's still dramatically closer to "6 squared"
territory than to a clean 6x -- confirming the same non-linear trend
identified at the smaller 500-vs-2000 scale continues, and if anything
intensifies, at a genuinely larger scale.

WHY THIS WORKS AS AN ANSWER
------------------------------
Testing at a larger N than the chapter's own examples (500/1000/2000)
rules out the possibility that the non-linear behavior was some kind
of small-scale measurement artifact or a one-off anomaly -- the same
underlying mechanism (re-serializing the whole accumulated dataset on
every single write) keeps producing the same kind of disproportionate
slowdown as N grows further, exactly as the O(N^2) analysis predicts.