Worklog #1628 — verify #932 logdet-channels build fix + fmt the relocated oracle

[SauersML]

Issue #1628 (5 build.rs ban violations from the #932 SAE logdet-channels revert 700e6c5) was already CLOSED — fixed on main by b316090 + 7abe885 (oracle relocated into #[cfg(test)] mod batch4_oracle_tests; #[allow], _n param, bare #[cfg(test)] gates all gone). I am NOT re-fixing — peers #1652/#1642/reopen-1614 own the adjacent jet-cutover build-break work (they touch row_jet_program.rs / multinomial_reml.rs / flex_jet.rs — disjoint from my file).

Verified at HEAD (156fe0a):

• ./build.sh with build.rs force-touched -> ban-scanner passes, BUILD OK.
• regression guard bug_hunt_sae_logdet_channels_revert_ban_violations_break_build -> 3/3 PASS.
• Taylor-jet scalar algebra: write each family's row log-likelihood once, derive the entire RowKernel derivative tower mechanically #932 oracle cross-check batch4_jet_lanes_match_scalar_hand_row_jets -> PASS (jet lanes match hand row_jets_for_logdet to 1e-9; oracle is genuinely exercised, not dead code).

Found + fixed a real leftover defect: the #932 relocation left the moved impl block UNDER-INDENTED (4-space bodies inside a nested mod+impl needing 8). On main, this file is rustfmt-clean EXCEPT exactly the 2 hunks of that relocated block, so it was authored fmt-conformant and the revert is the only drift. Applied rustfmt (edition 2024 per rustfmt.toml) -> file now fully clean; pure re-indent + benign width reflows, oracle test still passes. PR #1667.

[SauersML]

DONE — PR #1667. #1628 build break was already fixed on main; verified HEAD is fully green:

• ban-scanner passes (build.rs force-touched -> BUILD OK)
• 8/8 build-break ban-family regression guards pass
• Taylor-jet scalar algebra: write each family's row log-likelihood once, derive the entire RowKernel derivative tower mechanically #932 oracle cross-check passes (jet lanes match hand path to 1e-9)

Shipped fix: rustfmt the relocated batch4_oracle_tests impl block, which the #932 revert left under-indented (file was fmt-clean on main EXCEPT those 2 hunks). No semantic change, oracle test green post-fmt. Closing out my work on this issue.