fix: add fallback to polyfill when SIMD extensions are missing

[moefqy]

Problem

On macOS (specifically x86_64 / Rosetta environments), CPU architecture detection in deltakit_stim returns sse2 (or avx2), causing deltakit_stim/__init__.py to attempt:

if _tmp == 'avx2' or _tmp == 'sse2':
    from deltakit_stim._stim_sse2 import *

However, in the pre-built macOS wheel distribution on PyPI (deltakit_stim-0.2.0-cp311-cp311-macosx_15_0_x86_64.whl), _stim_sse2.so was omitted during the build, shipping only _stim_polyfill.so.

Because there is no exception handler around the import, installing and running deltakit on macOS raises an unhandled crash:

ModuleNotFoundError: No module named 'deltakit_stim._stim_sse2'

Solution

• Wraps the dynamic architecture import in a try...except (ImportError, ModuleNotFoundError) block.
• If _stim_sse2 is present, it will load for fast SIMD execution.
• If _stim_sse2 is missing from the installed package, it gracefully falls back to _stim_polyfill instead of crashing.

Verification

Tested on macOS 15 (Sequoia, Python 3.11.4):

• Confirmed import deltakit and import deltakit_stim now load cleanly without errors.
• Verified basic quantum circuit creation and sampling runs with 100% accuracy.

[CLAassistant]

All committers have signed the CLA.

[rolandriver]

Thank you @moefqy for this really. I've assigned reviewers to this PR and we will come back to you soon.