Allow re-running `panic`'ed tests through a runtime configuration flag

I'm using `gotestsum` to test a terraform provider that frequently has a very flaky acceptance test build (context in https://github.com/terra-farm/terraform-provider-xenorchestra/issues/188). The flakiness is due to tests becoming hung and eventually reaching `go test`'s timeout value. Re-running these tests results in a successful build, but requires a significant amount of manual work to identify what should be rerun.

I've tested a fork of `gotestsum` that allows rerunning `panic`'ed tests (https://github.com/ddelnano/gotestsum/commit/8c70ff05e83bc93e5e042151a4d25343f1dfe4f8) and it appears to work the way I intend it to -- tests that timeout are re-run and result in a passing test build. I created a [dummy test package](https://github.com/terra-farm/terraform-provider-xenorchestra/blob/2501b7d3e4076f88b87e09c511bf8f02fd9f1d89/cmd/testing/parallel/main_test.go) to test this. The test package includes 3 tests and will allow one of the tests to proceed with execution every 10 seconds. If `go test` is given a timeout greater than 30s all three tests pass. If the timeout is between 20s and 30s, then 1 of the tests will timeout and panic.

I used this package to test the behavior of my fork and verified that the result is expected.

```
# Verify that gotestsum re-runs the panic'ed tests and reports a successful build (exit code 0)
ddelnano@ddelnano-desktop:~/go/src/github.com/ddelnano/terraform-provider-xenorchestra$ ~/code/gotestsum/gotestsum --rerun-fails=2 --packages='github.com/ddelnano/terraform-provider-xenorchestra/cmd/testing/parallel' -- -v -count=1 -timeout=21s
✖ cmd/testing/parallel (21.016s)

DONE 3 tests, 1 failure in 21.846s

✓ cmd/testing/parallel (10.031s)

=== Failed
=== FAIL: cmd/testing/parallel TestParallelHangging (unknown)

DONE 2 runs, 4 tests, 1 failure in 32.883s

ddelnano@ddelnano-desktop:~/go/src/github.com/ddelnano/terraform-provider-xenorchestra$ echo $?
0

# Force test to always fail and verify that gotestsum reports a failure
ddelnano@ddelnano-desktop:~/go/src/github.com/ddelnano/terraform-provider-xenorchestra$ git diff
diff --git a/cmd/testing/parallel/main_test.go b/cmd/testing/parallel/main_test.go
index b47d341..1c9cfc9 100644
--- a/cmd/testing/parallel/main_test.go
+++ b/cmd/testing/parallel/main_test.go
@@ -38,4 +38,5 @@ func TestParallelHangging(t *testing.T) {
func TestParallelSlightHangging(t *testing.T) {
t.Parallel()
waitForChannel(c)
+ t.Errorf("Failed")
}

ddelnano@ddelnano-desktop:~/go/src/github.com/ddelnano/terraform-provider-xenorchestra$ ~/code/gotestsum/gotestsum --rerun-fails=2 --packages='github.com/ddelnano/terraform-provider-xenorchestra/cmd/testing/parallel' -- -v -count=1 -timeout=21s || echo $?
✖ cmd/testing/parallel (21.008s)

DONE 3 tests, 2 failures in 21.384s

✖ cmd/testing/parallel (10.014s)
✓ cmd/testing/parallel (10.012s)

DONE 2 runs, 5 tests, 3 failures in 42.515s

✖ cmd/testing/parallel (10.011s)

=== Failed
=== FAIL: cmd/testing/parallel TestParallelSlightHangging (20.01s)
main_test.go:41: Failed
panic: test timed out after 21s

goroutine 19 [running]:
testing.(*M).startAlarm.func1()
/usr/lib/go/src/testing/testing.go:2036 +0x8e
created by time.goFunc
/usr/lib/go/src/time/sleep.go:176 +0x32

goroutine 1 [chan receive]:
testing.tRunner.func1()
/usr/lib/go/src/testing/testing.go:1412 +0x4a5
testing.tRunner(0xc000116680, 0xc000106c78)
/usr/lib/go/src/testing/testing.go:1452 +0x144
testing.runTests(0xc0001300a0?, {0x605400, 0x3, 0x3}, {0x4a6be9?, 0x7f497f542fff?, 0x609ca0?})
/usr/lib/go/src/testing/testing.go:1844 +0x456
testing.(*M).Run(0xc0001300a0)
/usr/lib/go/src/testing/testing.go:1726 +0x5d9
github.com/ddelnano/terraform-provider-xenorchestra/cmd/testing/parallel.TestMain(0x7f49a6674a68?)
/home/ddelnano/go/src/github.com/ddelnano/terraform-provider-xenorchestra/cmd/testing/parallel/main_test.go:21 +0x5e
main.main()
_testmain.go:53 +0x1d3

goroutine 34 [sleep]:
time.Sleep(0x2540be400)
/usr/lib/go/src/runtime/time.go:195 +0x135
github.com/ddelnano/terraform-provider-xenorchestra/cmd/testing/parallel.TestMain.func1()
/home/ddelnano/go/src/github.com/ddelnano/terraform-provider-xenorchestra/cmd/testing/parallel/main_test.go:16 +0x2c
created by github.com/ddelnano/terraform-provider-xenorchestra/cmd/testing/parallel.TestMain
/home/ddelnano/go/src/github.com/ddelnano/terraform-provider-xenorchestra/cmd/testing/parallel/main_test.go:14 +0x54

goroutine 36 [chan receive]:
github.com/ddelnano/terraform-provider-xenorchestra/cmd/testing/parallel.waitForChannel(...)
/home/ddelnano/go/src/github.com/ddelnano/terraform-provider-xenorchestra/cmd/testing/parallel/main_test.go:25
github.com/ddelnano/terraform-provider-xenorchestra/cmd/testing/parallel.TestParallelHangging(0x0?)
/home/ddelnano/go/src/github.com/ddelnano/terraform-provider-xenorchestra/cmd/testing/parallel/main_test.go:35 +0x27
testing.tRunner(0xc000116b60, 0x531050)
/usr/lib/go/src/testing/testing.go:1446 +0x10b
created by testing.(*T).Run
/usr/lib/go/src/testing/testing.go:1493 +0x35f

=== FAIL: cmd/testing/parallel TestParallelHangging (unknown)

=== FAIL: cmd/testing/parallel TestParallelSlightHangging (re-run 1) (10.01s)
main_test.go:41: Failed

=== FAIL: cmd/testing/parallel TestParallelSlightHangging (re-run 2) (10.01s)
main_test.go:41: Failed

DONE 3 runs, 6 tests, 4 failures in 53.302s
1
```

While working on this, I realized that there is a history on handling this situation (https://github.com/gotestyourself/gotestsum/pull/192 and https://github.com/golang/go/issues/45508), so I know it was intentional that these tests aren't rerun but I'd like to make it possible to enable this behavior behind a new cli flag (`--rerun-fails-rerun-panics` or something similar).

I'm opening this issue to discuss if a PR that implements this would be accepted. If that is the case, I'm happy to take on this work myself.

Start with the behavior demonstrated in cmd/testing/parallel/main_test.go and review gotestsum/pull/192 plus golang/go issue 45508 for the existing decision about panic'ed tests. Trace how the rerun-fails option handles test output and statuses. Done means a new runtime flag can opt into rerunning timed-out tests while preserving the correct success or failure exit status.

Written by the indexing model from the issue text.