perf(react-icons-font-subsetting-webpack-plugin): improve plugin performance

[Hotell]

follows #961 (comment)

---

1) Regex handling + module filtering

• Before: isFluentUIReactFontChunk created and executed an inline RegExp for every module (/.../.test(m.resource)) . This was executed ~300 times while running npm test
• After: The RegExp is hoisted to a module-level constant (REACT_ICONS_FONT_MODULE_IMPORT_PATTERN) and isFluentUIReactFontChunk adds a cheap prefilter (resource.includes('react-icons')) before running the regex.

Impact: Fewer regex executions across large module graphs; lower per-module overhead.

2) Subset text generation complexity

• Before: Built subsetText by scanning all glyph entries in the font codepoint map:
  • Object.entries(codepointMap).filter(...).map(...).join('')
• After: Builds subsetText by iterating only usedExports and doing O(1) lookups in codepointMap:
  • for (const glyphName of usedExports) { codepointMap[glyphName] ... }

Impact: Changes from O(total glyphs in font) per font → O(used icons) per font, which is typically much smaller.

3) Correct async waiting (also affects perf/consistency)

• Before: await optimizationPromises; (does not actually await the array of promises).
• After: await Promise.all(optimizationPromises);

Impact: Ensures all subsetting work completes during optimizeAssets (fixes non-determinism and prevents leftover async work outside the compilation stage).

---

Perf gains

• about 0.10s faster (~4%) on our small synthetic test (npm test)
• a larger app/module graph should show the regex + glyph-scan wins more clearly.