Add performance benchmark CI to compare PR vs base branch (#32)

[shakayami]

Introduces a benchmarks/ suite (pytest-benchmark for execution time,
tracemalloc for peak memory) covering the computationally critical
modules: segtree, lazysegtree, dsu, maxflow, mincostflow, convolution,
and scc. A new benchmark.yml workflow runs the suite on both the PR
branch and the PR's base branch, diffs the results, and posts/updates
a single PR comment showing whether the change actually got faster or
used less memory.

Co-Authored-By: Claude Sonnet 5 noreply@anthropic.com
Claude-Session: https://claude.ai/code/session_016ZHPv3fCYTnGxKFpLVpnzn

[github-actions]

📊 Performance Benchmark Results

⏱️ Execution time (mean, lower is better)

Benchmark	Base	PR	Delta
convolution	n/a	986.44ms	n/a
dsu	n/a	111.86ms	n/a
lazysegtree	n/a	1.705s	n/a
maxflow	n/a	10.70ms	n/a
mincostflow	n/a	4.77ms	n/a
scc	n/a	178.68ms	n/a
segtree	n/a	254.18ms	n/a

💾 Peak memory (lower is better)

Benchmark	Base	PR	Delta
convolution	n/a	16.1MiB	n/a
dsu	n/a	849.2KiB	n/a
lazysegtree	n/a	5.0MiB	n/a
maxflow	n/a	2.3MiB	n/a
mincostflow	n/a	662.2KiB	n/a
scc	n/a	13.8MiB	n/a
segtree	n/a	1.5MiB	n/a

_{Base = target branch, PR = this branch. Negative delta = improvement. ✅ improvement ≥ 1%, ⚠️ regression > 5%. "n/a" base entries mean the target branch has no benchmark suite yet.}