Store energy-dependent isomeric production data verbatim in depletion chains

[jon-proximafusion]

Implements the chain-file side of #121: depletion chain files can now carry the evaluations' energy-dependent isomeric production data verbatim, so no side files are needed to know how a transmutation reaction splits between ground and metastable states.

What is stored

For each reaction target, an optional <isomeric_production> child element of <reaction> holding the ENDF MF=9 yield functions and MF=10 partial cross sections exactly as evaluated (original grids, values, and multi-region interpolation), together with the final-state level number, excitation energy, QM/QI, and the source library recorded in the evaluation header. No renormalization, thinning, or ratio pre-computation is performed; consumers fold the data with whatever spectrum they have. Old readers parse only the <reaction> attributes and ignore the child elements, and scalar branching ratios are unchanged by default (ground 1.0, metastables 0.0), so existing workflows including add_branching_ratios behave exactly as before.

Changes

• openmc/deplete/nuclide.py: ProductionTable and IsomericProduction dataclasses, a Nuclide.isomeric_production mapping keyed by (reaction type, target), XML round trip, and validation of the stored data (physical ranges, orphan entries, pointwise MF=9 sum checks). Also fixes a latent NameError in Nuclide.validate for nuclides with no decay modes and inconsistent reaction sums.
• openmc/deplete/_isomeric.py (new): MF=8/9/10 parsing from raw section text using the endf-python record readers, level-to-isomer assignment by excitation-energy matching against decay data (with positional pairing as fallback and fold-to-ground for unmappable levels, never naming targets from the LFS level index), scalar ratio collapse modes, and a mapping report.
• openmc/deplete/chain.py: Chain.from_endf gains isomeric_branching, branching_files (extra ENDF evaluations consulted where the primary neutron file has no MF=8/9/10 data, e.g. TENDL), scalar_branching, elis_rtol, and isomer_mapping_log arguments; reduce carries the data for retained targets; set_branch_ratios preserves attached data through scalar rewrites (with a preserve_isomeric_data safeguard); new get_isomeric_production accessor.
• Docs: chain XML format specification and Python API entries.

Verification

• About 40 new unit tests, including round-trip fixtures built from the verbatim ENDF/B-VIII.1 Am241 and Nb93 section text.
• Integration tests on the real ENDF/B-VIII.1 files reproduce the Am241 thermal capture split of 0.90 to ground and 0.10 to Am242_m1, with the metastable correctly mapped from level LFS=2 by excitation-energy matching (the issue that affected Adding isomeric branching to chain openmc-dev/openmc#3885).
• A full ENDF-B/VIII.1 + TENDL-2025 chain was built and round-tripped with this branch (numbers posted on Store energy-dependent isomeric branching data (MF=8/9/10) verbatim in depletion chain files #121: 31.7 MB vs the 27.7 MB baseline, 3,193 reaction-target pairs with data on 408 parents, built in 13 s) and loads and validates in unmodified upstream OpenMC 0.15.3.
• The branch was adversarially reviewed (four-dimension review with verification); confirmed findings, including a thermal-group dropping bug in the multigroup scalar collapse and a positional-mapping split of MF=9/MF=10 record pairs, were fixed in the final commit with regression tests.