Run each test in a separate process so that crashes don't stop the world.

[dabrahams]

Description

Forking is cheap, and this will allow a test suite to keep running even if one test unexpectedly fails. The Swift standard library tests work this way.

This requires giving up on "automatically parallelize tests in process," or at least allowing users to opt out, but not on scalable execution. Again, forking is cheap.

Expected behavior

No response

Actual behavior

No response

Steps to reproduce

No response

swift-testing version/commit hash

No response

Swift & OS version (output of swift --version && uname -a)

No response