Clean up Outdated Args from Llama Export Script

[YifanShenSZ]

2 coreml-specific args were introduced to llama export script

1. Partition Mutable Buffer as Core ML State #5165 introduced --coreml-enable-state
2. Preserve SDPA for CoreML #5258 introduced --coreml-preserve-sdpa

Since users would want to target iOS version rather than tune optimizations, #5319 introduced --coreml-ios 18. However, #5319 forgot to remove the outdated --coreml-enable-state and --coreml-preserve-sdpa

[pytorch-bot]

🔗 Helpful Links

🧪 See artifacts and rendered test results at hud.pytorch.org/pr/pytorch/executorch/5762

• 📄 Preview Python docs built from this PR

Note: Links to docs will display an error until the docs builds have been completed.

❗ 1 Active SEVs

There are 1 currently active SEVs. If your PR is affected, please view them below:

• PyTorch Testing Nodes Undergoing ROCm 6.2.1 Upgrades

✅ No Failures

As of commit 0b1e89c with merge base 905b88c ():
💚 Looks good so far! There are no failures yet. 💚

This comment was automatically generated by Dr. CI and updates every 15 minutes.

[YifanShenSZ]

No need to cherry pick this PR, since it won't affect any beta release functionality

[nil-is-all]

@YifanShenSZ Closing this PR since it has become stale. If you feel this is still relevant or you still face the issue discussed, feel free to re-open this PR.