Add a spec for stress testing garbage collection

[jordandcarter]

We ran into segmentation faults in production and eventually narrowed it down to selma. Here is a test which turns on GC.stress which reliably causes the segfault.

Issue raised here #81

[gjtorikian]

tl;dr for the fix: the memory for the hash sanitization config was initialized in Rust, then passed to Ruby, then used entirely in Rust. I had flagged that back and forth volleyball as an issue but it never caused an observable problem (until now).

I moved all the config setup entirely in Rust, and the test passes. 👍

Noticed some segmentation faults in production,

If Selma helps you at work, please, consider sponsoring. ✌️