Fix remaining v0->v1 conversion gaps (third body, gas phase, surface prefix)

[K20shores]

Follow-up to #309. That PR fixed the diffusion coefficient, but three more v0-parser issues block converting a real v0 (CAMP) config to a loadable v1 mechanism. (#309 was squash-merged with only its first commit, so the third-body and gas-phase fixes below — briefly part of that PR — are not actually in main; they are re-included here.)

Fixes

1. THIRD_BODY tracer dropped. A CHEM_SPEC with tracer type: THIRD_BODY recorded only the tracer-type string, which the unified type does not carry into v1 serialization (v1 uses is third body). Set is_third_body so the designation survives conversion — otherwise species like M silently lose their third-body role.

2. Empty reaction gas phase. v0 reactions left gas_phase empty, but the v1 parser requires every reaction to name a non-empty gas phase, so a converted mechanism failed to re-parse with Unknown phase ''. Place all v0 reactions in the gas phase, mirroring how all v0 species are placed there.

3. Doubled SURF. prefix. The v0 surface parser stored the name as "SURF." + MUSICA name — the only v0 reaction parser that prefixes the name (photolysis, emission, loss, user-defined all store the bare name and let the MICM conversion add the label prefix). Since that conversion also prepends SURF., surface labels became SURF.SURF.<name>, breaking consumers that build per-reaction columns from the name (e.g. music_box's SURF.<name>.<property>.<unit>). Store the bare name so the single downstream prefix yields SURF.<name>.

Tests

Extends the v0 species, arrhenius, and surface unit tests. All 12 v0 tests pass.

Together with NCAR/musica#983 (stamp v1 version on --convert), this lets musica-cli --convert produce a v1 mechanism that loads directly.

🤖 Generated with Claude Code