fix(physics): restore relative slingshot response

A full comparison against the 175-record ledger found stale guessed coordinates
in both relative slingshot chains. Circles 56/58/71/73 and lines 57/59/72/74
were displaced by up to 70 pixels. The corrected record-57 geometry then
exposed a second issue: readable C and Rust had collapsed two independent
binary response projections into one.

Transcribe the accumulated initializer coordinates exactly. Use the
swapped-component projection only to sign the current tangential response
`abs(normal_velocity) * response_tangent`, and use the standard dot product for
`spin_delta * 0.02 * response_tangent`. Correct the readable C evidence and
Rust together, extend live tracing with the six-byte spin field, and seal the
stopped-Wine transition from `(75530,354765)/(3000,15)` to
`(77369,356597)/(1839,1832)` with Real48 spin `-5.28` in both harnesses.

Test Plan:
- stopped Wine 11.15 record-57 injection/trace -- matched position, velocity, and spin
- `cargo test --workspace --all-targets --all-features` -- 123 passed
- `cargo clippy --workspace --all-targets --all-features -- -D warnings` -- passed
- `bash original/tools/test_reconstructed_c.sh` -- passed
- `python3 original/tools/audit_reconstruction.py --require-complete` -- passed with zero incomplete or unclassified units
- `python3 -m py_compile original/tools/trace_original_state.py original/tools/inject_original_state.py` -- passed
- `rumdl check --flavor commonmark RECONSTRUCTION.md CHANGELOG.md` -- passed
- `git diff --cached --check` -- passed

original/C_RECONSTRUCTION_FINAL_AUDIT.md

@@ -82,13 +82,16 @@ Raw `1000:531e` further corrects the Pascal-order BitBlt interpretation: the
 DAT994 progress strip is drawn at `(202,328)` with height 13 and width
 `(stage-1)*7`, rather than the impossible historical 328x202 read from a
 237x13 bitmap. The focused loading-stage harness seals stages 1, 35, and 0.
-Live one-substep rail/circle probes further correct the response projection:
-the binary tangent driver is the swapped-component value
-`-(vx*normal_y + vy*normal_x)/length`, multiplied by `response_tangent` and
-added to retained `spin * 0.4`; subtracting the bare coefficient incorrectly
-rounded it to zero.
-The documented horizontal drain transition `(0,3015)->(-302,-1809)` and the
-ordinary circle/bumper tangent transitions seal this correction.
+Live one-substep rail/circle probes further correct the response transform.
+The binary keeps two projections distinct: swapped-component
+`-(vx*normal_y + vy*normal_x)/length` supplies the sign of the tangential
+response magnitude `abs(normal_velocity)*response_tangent`, which is added to
+retained `spin*0.4`. Standard dot-product
+`(vx*normal_x + vy*normal_y)/length` independently drives
+`spin_delta = projection*0.02*response_tangent`. The documented horizontal
+drain `(0,3015)->(-302,-1809)`, ordinary circle/bumper transitions, and stopped
+Wine record-57 transition `(75530,354765)/(3000,15)` to
+`(77369,356597)/(1839,1832)` with exact Real48 spin `-5.28` seal both paths.
 
 ## Reproducible gates