ForkServerBuilder might reset config, causing hangs in persistent mode fuzzing

[bernhl]

Describe the bug
When using ForkServerBuilder.shmem_provider(), this resets some of the fields already configured. E.g., the kill signal is reset to its default value: https://github.com/AFLplusplus/LibAFL/blob/main/libafl/src/executors/forkserver.rs#L1326
When the child processes encounters a timeout, the child is killed with the default value (sigterm). This might be insufficient, an causes a hang in the following situation:

1. child executes a long-running input
2. fuzzer detects a timeout, about to send sigterm (sigkill was configured, but reset by shmem_provider to sigterm)
3. child issues the sigstop persistent mode signal
4. fuzzer requests new child from forkserver, tells forkserver that the child needs to be reaped bc it was killed
5. forkserver waitpids for child
6. because the child hasn't been terminated properly, the forkserver waitpids indefinitely
7. because the forkserver is not delivering a new child pid, the fuzzer waits indefinitely

[domenukk]

Nice find. Most forkserver executors have to use the Unix Shmem provider that's probably why nobody noticed so far.
I tried to fix it in #2540