FEAT: find transitions at quantum-number level

[grayson-helmholz]

• Closes Solve and collect transitions at the quantum-number level #387

✨ New features

• find_qn_transitions() solves a QNProblemSetCollection purely at the quantum-number level. It runs the CSPSolver without completing the intermediate states from a particle database, so each intermediate state carries exactly the quantum numbers that the problem set declares as domains. Its result is a tuple of QNTransitions, the new alias for a FrozenTransition of quantum-number property maps.
• collect_qn_transitions() is the counterpart of collect_reaction_info() for that workflow: it merges the initial facts into each solution and deduplicates the transitions, without consulting a particle database or constructing State objects.
• strip_spin_projections() removes the spin projections of the initial and final state, the spin-projection domains of the intermediate edges, and the $L$ and $S$ projection domains of the interaction nodes from a collection of problem sets. Problem sets that become identical, such as the expansion over all spin-projection combinations, are deduplicated. Together with find_qn_transitions() this yields $J^{P(C)}$-level transitions, as needed for a Dalitz-plot decomposition.
• strip_quantum_numbers() in qrules.solving is the general form of that operation for a single QNProblemSet. Conservation rules are deliberately kept: a rule that can no longer execute is skipped and reported through the solver's not-executed-rules mechanism, rather than silently disappearing.

⚙️ Enhancements

• qrules.io.asdict() serializes sets as sorted lists and maps callable dictionary keys to their name, recursing into nested mappings. This is what makes a QNProblemSet serializable at all, since its settings are keyed by conservation rule.

Squash commit messages

* ENH: serialize rule mappings and sets in qrules.io
* FEAT: add strip_quantum_numbers function
* FEAT: add strip_spin_projections function