Profiling: mechanism to check that signal handlers are still active

[danielsn]

Minor: Not sure this is something we want to care about, but one thing we could do when the function is called is to check that our signal handlers are still active, e.g. because something else on the system may have overwritten them.

(I'm thinking of this, because there may be a situation where someone starts the crash tracker, and then discovers that the VM overwrites their signal handlers after X, and then calls register_crash_handlers again expecting libdatadog to be restored but isn't. Maybe too farfetched?)

Originally posted by @ivoanjo in #297 (comment)