E2E testsuite producing false negatives after panics

[arminsabouri]

If the payjoin CLI or an end-to-end test panics, the test resources remain uncleared. Since env::temp_dir() might return the same temporary directory for subsequent test runs. Each test run is expected to start with a fresh slate, which is particularly important if the developer is modifying the data serialization format between tests.

A quick fix is to prefix the temporary directory with a timestamp to ensure uniqueness. A more comprehensive solution would be to implement a panic handler along with handlers for other signals (e.g., SIG_INT, SIG_KILL) that the test process might receive. Alternatively we could optimistically clean the temp dirs before we run the tests but to quote @DanGould "that smells". So I would advocate for the panic handler solution.

[DanGould]

Getting the test framework to be robust makes iteration faster. Panic handler to start sounds right. Not sure how other signals should be handled, or if there can be a catchall.

[arminsabouri]

if there can be a catchall.

I wouldn't do a catchall b/c some signals are normal and expected -- only handle signals that request a process to be terminated. For example, SIG_CHILD is expected here b/c of all the child processes we fork in the testsuite and we dont want to cleanup resources halfway through the test run