fix(arborist): prioritize valid workspace nodes

[nlf]

closes #3637

[npm-robot]

found 20 benchmarks with statistically significant performance improvements

• app-large: clean, lock-only, cache-only, cache-only:peer-deps, modules-only, no-lock, no-cache, no-modules, no-clean, no-clean:audit
• app-medium: clean, lock-only, cache-only, cache-only:peer-deps, modules-only, no-lock, no-cache, no-modules, no-clean, no-clean:audit

timing results

app-large	clean	lock-only	cache-only	cache-only peer-deps	modules-only	no-lock	no-cache	no-modules	no-clean	no-clean audit
npm@8	55.702 ±0.61	33.505 ±0.48	36.816 ±23.21	22.226 ±1.13	3.435 ±0.01	3.289 ±0.08	2.649 ±0.04	12.843 ±0.23	2.634 ±0.10	3.730 ±0.16
#4230	0.450 ±0.01	0.448 ±0.01	0.445 ±0.01	0.450 ±0.00	0.467 ±0.02	0.448 ±0.01	0.470 ±0.01	0.448 ±0.00	0.449 ±0.00	0.448 ±0.00

app-medium	clean	lock-only	cache-only	cache-only peer-deps	modules-only	no-lock	no-cache	no-modules	no-clean	no-clean audit
npm@8	40.896 ±0.95	25.372 ±0.06	14.180 ±0.02	15.631 ±0.03	3.048 ±0.04	3.091 ±0.12	2.727 ±0.02	9.799 ±0.16	2.436 ±0.02	3.332 ±0.10
#4230	0.457 ±0.02	0.476 ±0.03	0.460 ±0.01	0.475 ±0.01	0.457 ±0.01	0.446 ±0.00	0.460 ±0.00	0.464 ±0.02	0.460 ±0.02	0.459 ±0.03

[fritzy]

I wish this test had assertions to make it more explicit on the behavior you're looking for.

[ljharb]

It's specifically "the output is tab-indented" - other than ensuring that each line starts with either "not a space" or "a tab", i'm not sure what i'd do. what would you prefer?