perf: inline optional getArg in BasicSourceMapConsumer constructor

[7rulnik]

Summary

Counterpart to #62 — five util.getArg(sourceMap, X, default) sites in BasicSourceMapConsumer():

• names (default [] — Sass 3.3 deviation)
• sourceRoot (default null)
• sourcesContent (default null)
• file (default null)

Replaced with sourceMap.X != null ? sourceMap.X : default. Same pattern as #59 / #60 / #62.

Bench

Honest result: this one is a wash. Originally measured +1.1%; re-benched after rebase onto main (with #64/#65/#66 landed): mean −0.49% across 12 init-bench rows, range −3.8% to +3.1% — within fixture variance.

Unlike #62 (where the SourceMapGenerator ctor inline saw +8.7% on addMapping, likely from V8 hidden-class effects helping the downstream per-mapping hot loop), the consumer ctor isn't followed by a hot loop in the same bench cycle. _parseMappings dominates init cost, so saving four function calls in the preamble is unmeasurable.

Keeping the change anyway for codebase consistency — without it, a future reader of #62 would have to wonder why one constructor was inlined and the other left alone.

SOLO=1 PHASES=init scripts/bench-diff.sh main trace (two-process, ops/sec via benchmark.js, after rebase):

Fixture	JSON cand	JSON base	JSON Δ	Object cand	Object base	Object Δ
amp.js.map	521 ±0.39%	527 ±0.44%	-1.1%	617 ±0.43%	641 ±0.29%	-3.7%
babel.min.js.map	78.03	78.52	-0.6%	91.38	89.74	+1.8%
issue-41.js.map	112 ±1.41%	112 ±1.59%	0.0%	127 ±0.92%	127 ±0.85%	0.0%
preact.js.map	9,340 ±0.23%	9,460 ±0.40%	-1.3%	18,566 ±0.51%	19,292 ±0.30%	-3.8%
react.js.map	5,605 ±0.59%	5,691 ±0.34%	-1.5%	6,727 ±0.27%	6,680 ±0.42%	+0.7%
vscode.map	5.40 ±5.72%	5.24 ±6.00%	+3.1%	7.15 ±5.58%	7.11 ±9.18%	+0.6%
mean						-0.49%

Tests

• All 205 existing tests pass.
• yarn test:coverage thresholds (98/97/96) green:
  • lib/source-map-consumer.js: 99.93 / 98.34 / 97.30
  • all files: 98.35 / 97.21 / 96.75
• No new tests needed — the existing fixture set already exercises both branches of each default ternary, so coverage holds on its own.

[7rulnik]

Re-benched after rebase onto current main (with #64 / #65 / #66 landed). Result holds: mean −0.49% across 12 init rows (range −3.8% to +3.1%) — wash, within fixture variance. Original conclusion unchanged. Pushed rebased branch to update the diff.