Fix/issue 1282 mypy batch3 part2

[rosspeili]

Fix mypy errors in batch 3 part 2 (#1282)

Typing-only, no runtime changes.

• hamiltonians/special_operators.py, majorana_operator string parsing types, coefficient parameter annotation, number_operator return/Op types
• contrib/representability/constraints/spin_orbital_2pos_constraints.py, fix incorrect Optional on g2d2map factor
• contrib/representability/_dualbasis.py, correct scalar/bias/tensor_elements types (follow-imports dependency for spin_orbital constraints)

hartree_fock.py remains deferred, will do in next PR.

Test plan

• strict mypy (ignore_errors=false, full follow-imports) on all 3 files
• pytest on special_operators_test.py, spin_orbital_2pos_constraints_test.py, _dualbasis_test.py
• black on changed files

[gemini-code-assist]

Code Review

This pull request modernizes type annotations across several files by adopting PEP 604 union types (|) and standard collection types. The review feedback correctly points out that because the project targets Python 3.9, evaluating these new union types at runtime will raise a TypeError unless from __future__ import annotations is imported. The reviewer suggests adding this future import, simplifying redundant union types (e.g., using float instead of float | int), and annotating function parameters directly to eliminate the need for runtime typing.cast calls which would otherwise fail.

[rosspeili]

Pushed a small hotfix based on suggestions, annotating coefficient: int | float | complex on majorana_operator and removed the typing.cast, but without the other changes.

Not applying the rest of the gemini suggestions:

• from __future__ import annotations, OpenFermion requires Python 3.10+, we already use X | Y unions elsewhere (eg. plane_wave_hamiltonian) without __future__. The Python 3.9 concern doesn't apply here.
• factor: float instead of float | int on g2d2map, cosmetic only, so keeping float | int for consistency with _dualbasis and prior Mypy reports many errors, and they should be fixed #1282 PRs.

Mypy + tests still pass locally on all 3 files. @mhucka let me know if I am missing something or if I should apply other suggestions too.

[mhucka]

This looks good! Thank you.