iOS 18 Target Regression

## 🐞Describing the bug
- My model works as expected when converting to `coremltools.target.iOS17`, but only outputs `nan` when targeting `coremltools.target.iOS18`

## To Reproduce
- Clone this repository: https://github.com/NoahBPeterson/mxbai-embed-large-v1-CoreML (commit 4121f6d5ab6506ba7c5a18cac320ab92f9b6541f)
- Run `convert_pytorch_coreml.py`.
- Test the output using `run_coreml.py`
- It should output an array of floats.
- Change line 40 in `convert_pytorch_coreml.py` to `minimum_deployment_target=ct.target.iOS18,`
- Convert the model again
- Run: the model only outputs `nan`

## System environment (please complete the following information):
- coremltools version: 8.0
- OS (e.g. MacOS version or Linux type): macOS 15.0
- Any other relevant version information (e.g. PyTorch or TensorFlow version): torch 2.4.0, Python 3.11

Start by cloning the linked mxbai-embed-large-v1-CoreML repository at commit 4121f6d5ab6506ba7c5a18cac320ab92f9b6541f and running convert_pytorch_coreml.py followed by run_coreml.py. Compare the iOS 17 and iOS 18 targets around line 40; done means the iOS 18 conversion produces an array of floats rather than nan.

Written by the indexing model from the issue text.