New vfs baselines follow-up

[rbuckton]

Per #20763 (comment), we opted to "patch" some of the paths used in the VFS PR to reduce the size of the PR and accept the baseline changes later. These are the baseline changes resulting from removing these "patches".

[mhegazy]

this will use lib.d.ts by default for tests, we opted to make the lib file smaller in tests to make sure we are not rechecking lib.d.ts all the time. what impact does this have on the test runtime?

[weswigham]

Why do so many tests now have an explicit // @lib: es5?

[rbuckton]

They contain declarations that conflict with declarations in dom. For those cases it was easier to just set the file to use lib.es5.d.ts.

[rbuckton]

I ran Measure-Command { gulp runtests-parallel } against master and oddly enough this was faster:

C:\dev\TypeScript [master]> gulp clean
...
C:\dev\TypeScript [master]> gulp tests
...
C:\dev\TypeScript [master]> Measure-Command { gulp runtests-parallel }
...
TotalSeconds      : 310.9653838
...


C:\dev\TypeScript [vfsBaselines]> gulp clean
...
C:\dev\TypeScript [vfsBaselines]> gulp tests
...
C:\dev\TypeScript [vfsBaselines]> Measure-Command { gulp runtests-parallel }
...
TotalSeconds      : 300.0299655
...

[weswigham]

oddly enough this was faster

AFAIK we don't actually check the std libs we inject (unless added via the libFiles psuedo-compileroption) - we pull whatever's necessary to check the test. This PR removes a munch of errors from older baselines (by adjusting the lib they use); errors take time to generate, emit, and compare, so I'd wager that's where the difference lies.