Fix crash on extension reload

[lilleyse]

Creating a Fabric prim with the same path as a previously destroyed prim causes a crash. Previously this was avoided by including the object id and pool id in the name, but on extension reload the pool id is initialized to 0 and new paths may conflict with old paths.

Now a contextId is randomly generated and ensures that Fabric prim paths are unique even across extension reloads.

There are probably more robust and safer ways to do this than glm::linearRand<int64_t>(0, 100000), like using a timestamp, but since extension reloads are uncommon and I didn't want to overcomplicate this.

With the fix applied the extension no longer crashes when disabled and re-enabled:

extension-reload.mp4

[corybarr]

Simple and clean.

My goto-for these issues has been a timestamp like you mention in a one-liner. I don't think it would add any real amount of complexity if you were to drop it in. But the odds of a name-collision crash are also insignificant.

[lilleyse]

I switched to a timestamp; it's a better solution.