Initial

[sanchda]

What does this PR do?
libdatadog v14.0.0 ships a new release of crashtracking, which makes the component safer for a wider variety of applications. At a high level, the main design difference is that crashtracking now waits for a crash to spawn a receiver process, and it tries very hard to ensure that process is reaped by the crashing process.

• Change the top-level crashtracking timeout from seconds to milliseconds (and improve the implementation of that timeout to be more strict)
• Remove init functions--now there's just one
• Remove "wait for receiver" as this is now implied
• Decouple altstack creation from altstack usage

Motivation:
Hopefully this fix makes crashtracking better for Ruby

[ivoanjo]

Thanks @sanchda for starting the libdatadog 14 upgrade! It got finished in #4065, so closing this one for now :)