Tests are grouped.

t/0*: bootstrap: if these fail, bail out and don't waste CPU
t/1*: core modules in isolation (marked with module name)
t/2*: refute{ ... } mini-apps for most common functionality
t/3*: extra check libraries, experimental features
t/4*: failure mode testing: check warnings, exceptions, and error messages
t/5*: deprecated features (marked with EOL version/date)
t/6+: interoperation with other systems

TODO bootstrapping sequence:
1) Verify refute(), is_passing(), get_tap(), and get_sign()
2) Verify try_refute { ... }
3) Verify is, like assertions
4) Verify contract_is, test_test assertions
5) Verify working in a harness using fork & pipe

