pkg/repro: tolerate scp errors

[a-nogikh]

We see numerous failures like this:

message : repro failed: failed to copy syz-execprog to VM: failed to run ["scp" "-P" "22" "-F" "/dev/null" "-o" "UserKnownHostsFile=/dev/null" "-o" "IdentitiesOnly=yes" "-o" "BatchMode=yes" "-o" "StrictHostKeyChecking=no" "-o" "ConnectTimeout=10" "-v" "/syzkaller/syzkaller/current/bin/linux_amd64/syz-execprog" "root@IP:./syz-execprog"]: exit status 255

It should okay to just retry the reproduction step several times rather than abort the whole process -- in the end, we're fuzzing instrumented developmental trees and there can be all sorts of sporadic problems.

[a-nogikh]

syzkaller/pkg/repro/repro.go

Lines 593 to 603 in 5e7b4bc

	const attempts = 3
	for i := 0; i < attempts; i++ {
	// It's hard to classify all kinds of errors into the one worth repeating
	// and not. So let's just retry runs for all errors.
	// If the problem is transient, it will likely go away.
	// If the problem is permanent, it will just be the same.
	result, err = callback()
	if err == nil {
	break
	}
	}

Ah, we actually already do it 3 times. So for some kernels scp fails 3 times in a row?