std tests: avoid duplicating thread-local machinery

[RalfJung]

Under cfg(test) we have two copies of the standard library loaded, including two copies of all its global variables. Something about how that interacts with #148799 results in memory leaks on windows-gnu targets, which are caught by Miri.

I'm not sure if it's worth debugging this. It's easier to just avoid duplicating the thread-local machinery.

(This is something we had already done in the past, and then at some point stopped doing as Miri showed no more problems with the duplicate machinery. Now the problems are back.)

Cc @joboet @ChrisDenton

[rustbot]

r? @clarfonthey

rustbot has assigned @clarfonthey.
They will have a look at your PR within the next two weeks and either review your PR or reassign to another reviewer.

Use r? to explicitly pick a reviewer

Why was this reviewer chosen?

The reviewer was selected based on:

• Owners of files modified in this PR: @ChrisDenton, libs
• @ChrisDenton, libs expanded to 10 candidates
• Random selection from Mark-Simulacrum, clarfonthey, jhpratt

[clarfonthey]

Duplicating libstd in tests was probably bound to do something like this eventually, honestly. The code looks fine and I guess while it's tempting to find a permanent fix I feel more inclined to just merge as a direct improvement.

r=me from the libs side but will let you tell bors to merge in case you wanted more input on the review first. Your comment implies no, but since this is sufficiently weird and you have r+ permissions anyway I think it's fine to let you do it.