test(libnpmpublish): unpublish from custom reg

[ruyadorno]

Adds a test that ensures it's possible to unpublish a specific version
of a package from a custom registry (containing multiple slashes)

References

Relates to: #4253

[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	50.710 ±0.29	35.626 ±0.34	26.903 ±7.11	24.606 ±0.93	3.434 ±0.03	3.487 ±0.19	2.884 ±0.05	15.174 ±0.26	2.848 ±0.01	3.992 ±0.10
#4484	0.519 ±0.01	0.547 ±0.01	0.532 ±0.00	0.507 ±0.01	0.516 ±0.01	0.500 ±0.01	0.508 ±0.02	0.491 ±0.01	0.507 ±0.02	0.515 ±0.02

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	38.881 ±0.24	27.484 ±0.16	16.563 ±0.32	17.433 ±0.01	3.464 ±0.01	3.419 ±0.06	2.985 ±0.12	11.181 ±0.26	2.788 ±0.05	3.760 ±0.02
#4484	0.502 ±0.01	0.550 ±0.05	0.532 ±0.01	0.513 ±0.03	0.558 ±0.01	0.528 ±0.00	0.512 ±0.01	0.510 ±0.00	0.542 ±0.02	0.505 ±0.01

[ruyadorno]

Continued here: #4657