minifier: Performance regression: 2.5x slower minification with minor input size increase

[IWANABETHATGUY]

Description

I've encountered a significant performance regression in oxc-minify where a file that is only ~6% larger takes 2.5x longer to minify.

Benchmark Results

File	Input Size	Output Size	Compression Ratio	Time
good.js	10176.62 KB	6781.65 KB	33.36%	282.31 ms
bad.js	10820.12 KB	6937.99 KB	35.88%	704.10 ms

Environment

• oxc-minify version: 0.105.0
• Platform: Windows (win32-x64-msvc)

Reproduction

1. git clone https://github.com/IWANABETHATGUY/oxc-minify-performance-issue
2. pnpm install

Files attached: good.js and bad.js

node minify-with-oxc.mjs ./good.js  # ~282ms
node minify-with-oxc.mjs ./bad.js   # ~704ms

Expected Behavior

Similar input sizes should result in similar minification times.

Actual Behavior

A 6% increase in input size causes a 150% increase in processing time, suggesting possible algorithmic complexity issues with certain code patterns.

---

[armano2]

looks like the difference in code is var import_react$9947 = /* @__PURE__ */ __toESM(require_react()); * 10000, that is a 10k root level variables

looks like in "bad" version we also do not remove calls -> var import_react$9947 = /* @__PURE__ */ __toESM(require_react()) becomes require_react()