Closes #4.

Adds a small interpreted layer over the existing MF=8/9/10 parsers, as proposed in #4.

What this adds

New module src/endf/radionuclide_production.py:

• RadionuclideProduction dataclass: production data for a single final state of one reaction, joining the MF=8 identification of a radioactive product (ZAP, LFS level number, excitation energy) with the MF=9 yield multiplicity and/or MF=10 production cross section for that state. An excitation_energy property takes the MF=8 ELFS when present and otherwise falls back to QM - QI.
• radionuclide_production(material) function: for every reaction that has an MF=9 or MF=10 section, returns the final states in evaluation order, merging the MF=9 and MF=10 data for the same (ZAP, LFS) pair and attaching the MF=8 excitation energy when available. Returns a mapping of MT numbers to lists of RadionuclideProduction.

LFS is a level index of the product nuclide, not an isomeric-state index; mapping a level to a metastable state requires comparing excitation_energy against decay data. This is recorded in the docstrings.

Both names are exported from the top-level endf package.

Tests

New tests/test_radionuclide_production.py (5 tests) with a 46 kB trim of the ENDF/B-VIII.1 In-115 evaluation as a fixture (tests/n-049_In-115_trimmed.endf). It covers the three interesting shapes in one file:

• MF=9 yields with an implicit ground-state share (capture to In116 m1, Y = 0.79)
• MF=10-only threshold states with no MF=9 counterpart (MT=4 and MT=16)
• excitation energy attachment from MF=8

The full test suite (21 tests) passes.