ltm: surface MostlyReinforcing/MostlyBalancing in JSON/TS/WASM layers #495

Problem

PR #490 added MostlyReinforcing/MostlyBalancing (Rux/Bux) variants to the Rust LoopPolarity and DetectedLoopPolarity enums, plus a polarity_confidence: f64 field on DetectedLoop. The JSON-emitted to_loop_summary (in src/simlin-engine/src/analysis.rs) surfaces the new variants as mostly_reinforcing / mostly_balancing strings.

However, the TypeScript / WASM / C-FFI layers still see only the three legacy values:

• src/libsimlin/src/ffi.rs SimlinLoopPolarity only has Reinforcing = 0, Balancing = 1, Undetermined = 2. src/libsimlin/src/analysis.rs coalesces MostlyReinforcing -> Reinforcing and MostlyBalancing -> Balancing before returning to FFI consumers, with the polarity confidence dropped on the floor.
• The @simlin/engine / @simlin/diagram packages downstream of the WASM bindings therefore have no way to distinguish a steady-R loop from a Rux loop, even though the engine knows the difference.

This was an accepted trade-off for #490 to keep the FFI surface stable, but it's tech debt worth tracking separately.

What needs to happen

1. Extend the C ABI: either add Rux = 3, Bux = 4 to SimlinLoopPolarity plus a polarity_confidence: f32 field on SimlinLoop, or expose confidence through a new accessor. Whichever shape is chosen needs to be additive so existing callers keep compiling.
2. Update the cbindgen-generated simlin.h and the WASM @simlin/engine types to match.
3. Plumb the new variants through src/diagram/ so the visual rendering can distinguish Rux / Bux (the LTM literature suggests a "gray" treatment, similar to the confidence < 0.99 rendering in section 13.7 of docs/reference/ltm--loops-that-matter.md).
4. Mirror the FFI extension into pysimlin so its LoopPolarity enum can survive a round-trip through the FFI without coalescing the way the Python-side from_runtime_scores already preserves the variants.

Context

Discovered during PR #490 review (#490). The PR body acknowledged the FFI coalescing as a deliberate scope cut; this issue records the follow-up.

References

• src/libsimlin/src/analysis.rs (current coalescing site)
• src/libsimlin/src/ffi.rs (SimlinLoopPolarity)
• src/simlin-engine/src/db_analysis.rs (DetectedLoopPolarity, polarity_confidence)
• src/simlin-engine/src/ltm.rs (POLARITY_CONFIDENCE_THRESHOLD, LoopPolarity)
• docs/reference/ltm--loops-that-matter.md section 13.7 (formula and gray-rendering convention)
• LTM tracking epic: epic: Loops that Matter (LTM) -- open work #488