Accept numpy/int scalar temperatures in Mixture.calc_property

[djkees]

Summary

Mixture.calc_property's temperature argument dispatch only recognized Python float as a scalar temperature. NumPy floating scalars (e.g. np.float32, which does not subclass float) fell through to the method's else branch and raised a spurious ValueError, even though such a value arises naturally from indexing into a float array (arr[0]). Plain Python int was rejected for the same reason.

Changes

• source/bind/python/CEA.pyx: both scalar-dispatch checks in Mixture.calc_property (the pressure-dependent and pressure-independent branches) widened from isinstance(temperature, float) to isinstance(temperature, (float, int, np.floating)). Plain NumPy integer scalars (np.int64, etc.) remain deliberately rejected.
• source/bind/python/tests/test_mixture.py: 5 new tests covering the dispatch boundary — numpy scalar (float32/float64) accepted and matching the float result, plain Python int accepted and matching the float result, numpy integer scalar still rejected, and an invalid type (string) still rejected.

Testing

• make py-rebuild
• pytest source/bind/python/tests -v — 118 passed
• Manually reproduced the original repro (np.float32 scalar temperature); confirms it now returns a result instead of raising ValueError

Compatibility / Numerical behavior

• No expected changes to numerical results

Only the type-dispatch logic changed — no change to what value is ultimately passed into the Fortran solver for any input that was already accepted.

---

Drafted with Claude's assistance

• Confirmed the original bug empirically (isinstance(np.float32(3.0), float) is False while isinstance(np.float64(3.0), float) is True on this platform, explaining why only some numpy scalar dtypes triggered it) before writing the fix.
• Ran the full Python test suite after rebuilding the extension (make py-rebuild) against this branch's upstream/main base and confirmed all 118 tests pass, including the 5 new ones added for this fix.

🤖 Generated with Claude Code

[djkees]

Superseded by #163, which combines this fix with #161's into one PR since both touch the same code region.