[zhiaos]

Companion to webgestalt_rust#36, which removes MetaAnalysisMethod::Fisher — its implementation returned a chi-square density rather than a p-value (#30).

Merge this first. It drops the package's dependency on that variant, so the next re-vendor after #36 lands compiles cleanly.

No behaviour change

multiOraEnrichment.R passed "fisher" to rust_multiomics_ora, and then discarded the meta-p it got back and recomputed it with poolr::stouffer a few lines later. multiswGsea.R does the same for GSEA. So the Fisher path ran on every multi-omics ORA analysis and its output never reached a user — no published result is affected.

It now asks for "stouffer", which is what the R layer was doing anyway.

Changes

• src/rust/src/lib.rs — both rust_multiomics_ora and rust_multiomics_gsea selected Fisher on the string "fisher" and Stouffer on anything else; both now select Stouffer unconditionally. The second site is easy to miss: it sits inside the combo_method == "meta" branch.
• R/multiOraEnrichment.R — passes "stouffer".
• The method_modifier roxygen no longer offers "fisher" as a choice.

The vendored webgestalt_lib still has the variant, so this compiles as-is; it simply stops naming it.