use nextest for better test isolation

[pompon0]

it will allow us to set panic=abort per test (because each test is executed in a separate subprocess) and set test timeouts without developing a custom test harness. Additionally I've set the default test timeout to 2min (period = 1min, killing test after 2 periods), which seems reasonable, given that all presubmit tests execute <1min as of today.

[pugachAG]

+1 for enabling this to set timeout for individual unit tests.

I was debugging flaky peer_manager::tests::routing::archival_node test and started preparing a PR to set timeout for cargo test using RUST_TEST_TIME_UNIT env variable along with -Z unstable-options --ensure-time, but this seems like a nicer solution!

[nagisa]

I would argue that we want to adjust documentation to inform developers to use nextest locally as a followup as well. Using different test runners can and will lead to weird hard to diagnose issues (why is it passing on CI but not locally?) eventually.

[akhi3030]

I would argue that we want to adjust documentation to inform developers to use nextest locally as a followup as well. Using different test runners can and will lead to weird hard to diagnose issues (why is it passing on CI but not locally?) eventually.

+1. Please add some documentation in https://near.github.io/nearcore/practices/testing/index.html (https://github.com/near/nearcore/tree/master/docs/practices/testing)

[pompon0]

(why is it passing on CI but not locally?)

We have that already, even though we are using the same test harness in CI and locally. Updated documentation.