Various fixes for invalid things

[dzaima]

This fixes a couple things:

• delete checkpoint without an argument reads outside array bounds

• Checkpoints created at the end state, or when the process has finished, cannot be functionally restored, so disallow those

• After a failed run, gdb generally doesn't query anything, except info threads (perhaps there are more, but I haven't looked). And if rr responds with any threads, things soon break. Perhaps there's a better way to check for/handle this than my introduced bool failed_restart;?

[khuey]

How does gdb ever get fixed once failed_restart is set to true?

[dzaima]

Oh is that why thread listing started sometimes not showing threads beyond the active one.. looking into it.

[dzaima]

Should be good now. failed_restart being unset at the same place as in_debuggee_end_state makes sense.

[rocallahan]

Can you fold your fixes into the original commits they modify?

[dzaima]

Rebased to latest, and added a fix of a new problem which happens to already get caught by two of my modified tests here - a simple c; c; when (or any rr-cmd) resulted in a python error.

[rocallahan]

Thanks!