[Bug]: Another OCIO related crash

### What happened?

I'm now experiencing another OCIO related crash in both commercial RV-2024.2.0 and OpenRV (latest) under MacOS. It always crashes in an IPGraph Eval thread and looks like a FrameBuffer member is accessed after the FrameBuffer instance has been deleted. I can prevent the crash if I disable caching using the "-nc" commandline option. Attached is a repro with instructions on how to reproduce.

[rvocio_crash.zip](https://github.com/user-attachments/files/18484684/rvocio_crash.zip)

### List all the operating systems versions where this is happening

macOS 15.2

### On what computer hardware is this happening?

MacBook Pro (Apple M1 Pro)

### Relevant console log output

```shell
Here is the most common stacktrace:

* thread #46, name = 'IPGraph Eval #2', stop reason = EXC_BAD_ACCESS (code=1, address=0x37e8e87d8b480f04)
* frame #0: 0x0000000112f3aa9a libTwkFB.dylib`TwkFB::FrameBuffer::identifier() const + 26
frame #1: 0x0000000112f3ab79 libTwkFB.dylib`TwkFB::FrameBuffer::identifier() const + 249
frame #2: 0x0000000112f3ab79 libTwkFB.dylib`TwkFB::FrameBuffer::identifier() const + 249
frame #3: 0x0000000112f3ab79 libTwkFB.dylib`TwkFB::FrameBuffer::identifier() const + 249
frame #4: 0x0000000112f3ab79 libTwkFB.dylib`TwkFB::FrameBuffer::identifier() const + 249
frame #5: 0x0000000112f3ab79 libTwkFB.dylib`TwkFB::FrameBuffer::identifier() const + 249
frame #6: 0x0000000100156718 RV`IPCore::Shader::operator<<(std::__1::basic_ostream>&, IPCore::Shader::ImageOrFB const&) + 152
frame #7: 0x00000001001649f0 RV`IPCore::Shader::Expression::outputHash(std::__1::basic_ostream>&) const + 256
frame #8: 0x00000001001b63e3 RV`IPCore::IPImage::computeRenderIDs() const + 691
frame #9: 0x00000001001b7938 RV`IPCore::IPImage::computeRenderIDRecursive() + 40
frame #10: 0x00000001001b7938 RV`IPCore::IPImage::computeRenderIDRecursive() + 40
frame #11: 0x00000001001b7938 RV`IPCore::IPImage::computeRenderIDRecursive() + 40
frame #12: 0x00000001001b7938 RV`IPCore::IPImage::computeRenderIDRecursive() + 40
frame #13: 0x00000001001b7938 RV`IPCore::IPImage::computeRenderIDRecursive() + 40
frame #14: 0x00000001001cfae0 RV`IPCore::IPGraph::evaluate(int, IPCore::IPNode::ThreadType, unsigned long) + 256
frame #15: 0x00000001001d057a RV`IPCore::IPGraph::evalThreadMain(IPCore::IPGraph::EvalThreadData*) + 1130
frame #16: 0x00000001001c9320 RV`IPCore::evalThreadTrampoline(IPCore::IPGraph::EvalThreadData*) + 272
frame #17: 0x000000010b3709a7 libstl_ext.dylib`stl_ext::thread_group::worker_jump() + 791
frame #18: 0x000000010b37040d libstl_ext.dylib`stl_ext::thread_group::thread_main(void*) + 173
frame #19: 0x00007ff80935c253 libsystem_pthread.dylib`_pthread_start + 99
frame #20: 0x00007ff809357bef libsystem_pthread.dylib`thread_start + 15
```

### Environment variables

_No response_

### Extra information

_No response_

### Code of Conduct

- [x] I agree to follow this project's Code of Conduct

Start by following the instructions in rvocio_crash.zip and reproduce the crash on macOS with caching enabled, then compare it with the -nc behavior. Use the IPGraph Eval stack trace, especially FrameBuffer::identifier(), Shader::Expression::outputHash(), and IPImage::computeRenderIDs(), to investigate the invalid FrameBuffer access; done means the supplied reproduction no longer crashes with caching enabled.

Written by the indexing model from the issue text.