added multiline output keyword support; propogated ions into product …

[markleader]

Summary

Fix two legacy CLI compatibility issues affecting CEARUN-style inputs:

1. Prevent EqSolver_init element-list mismatch crashes for ionized legacy problems by ensuring reactants/products are initialized with consistent ions handling.
2. Restore legacy cumulative parsing behavior for repeated output datasets (multiple output ... lines in one problem).

Changes

• Updated product mixture construction to propagate ions=prob%problem%include_ions in:
  • run_thermo_problem (source/main.f90)
  • run_shock_problem (source/main.f90)
  • run_deton_problem (source/main.f90)
• Updated input parser dataset read loop in source/input.f90:
  • When currently parsing an outp dataset, consecutive outp keyword lines are merged into the same buffer instead of starting a new dataset that overwrites prior output options.
  • This matches legacy cea2.f behavior for repeated output directives.
• Added parser regression coverage in source/input_test.pf:
  • New test test_parse_outp_repeated_dataset_keywords
  • Verifies repeated outp tokens cumulatively set mass_fractions, transport, trace, and preserve siunit.

Testing

• Manual legacy CLI runs in build-dev-ub-hunt/source:
  • ./cea cearun -t ../../data/thermo.lib -r ../../data/trans.lib
  • ./cea example1 -t ../../data/thermo.lib -r ../../data/trans.lib
• Output/reference check:
  • diff -u cearun-ref.out cearun.out (verified crash is resolved and expected output sections are present; formatting/style differences remain vs legacy reference formatting)
• Full tests:
  • cd build-dev-ub-hunt && ctest --output-on-failure (12/12 passed)

Compatibility / Numerical behavior

• No expected changes to numerical results
• Expected changes (explain and provide validation)

Expected behavior change: ionized legacy CLI cases that previously aborted with
CRITICAL: eqsolver_init: Element lists for products and reactants must match.
now run to completion. No solver algorithm changes were introduced; updates are limited to initialization consistency and legacy-compatible output parsing.