allow tests to fail

[andrewrk]

I propose to modify the default test runner so that when a test returns an error, the test is marked as failure, however subsequent tests are still run. At the end will be reported the failures and successes, and the final exit code of zig test will be 0 if and only if all tests succeeded with no memory leaks.

Follow-up proposal: modify std.testing.expectEqual and similar functions to return error.ZigTestFailed rather than calling panic.

Motivation is to increase developer iteration speed; it might help to find out more than one thing is broken at a time.

[marler8997]

Maybe provide an option to quit on the first failure as well?

[kubkon]

This will be a good first issue once it is accepted.

[daurnimator]

Follow-up proposal: modify std.testing.expectEqual and similar functions to return error.ZigTestFailed rather than calling panic.

This implies that all call-sites will now need to be updated to do try std.testing.expect*?

[ityonemo]

Yes please. I think I had questions about how to recover from a panic that I won't need when hooking up zig tests into elixirs test suite.