IBSEB: rename the ibseb_advance timestep parameter, which shadowed ERF::dt

[hgopalan]

What

ERF::ibseb_advance takes a dt parameter that shadows ERF's own dt member (amrex::Vector<double> dt, ERF.H:1165). A clang -Wshadow build warns at Source/ImmersedBoundarySEB/ERF_IBSEB.cpp:131, and the HIP and GCC CI jobs build with -Wshadow over IBSEB, which is compiled unconditionally.

Renamed to dt_lev: the name ERF::Advance uses for the same quantity, and the name the only caller already passes. Declaration, definition and the two uses in the body. Nothing else.

Found by the -Wshadow build run while checking #432; it is not from that PR.

Checks

• Fails before, clean after. The same build on the pre-fix tree warns declaration shadows a field of 'ERF' [-Wshadow] at ERF_IBSEB.cpp:131; on this branch: no warnings or errors outside Submodules (176 s, Debug, MPI, dust on, unit tests on).
• Tests: 681 unit tests pass; ctest -R IBSEB 5 of 5 pass, including the IBSEB_Cube regression (58.5 s) and the three Louis gtests
• Results unchanged: IBSEB_Cube fcompares against its gold file and passes, so the fields are unchanged. The rename touches no expression.
• Also run: codespell, git diff --check, the plan-file prose grep, the MSVC scan (long, M_PI, gmtime) on the changed file, the device-code scanners (no hits), Sphinx (zero warnings), submodule gitlinks all 160000.
• Not run, with reasons: no CUDA or HIP compile here (the PR's GPU jobs are the check); no single-precision build and no fire-OFF build, since a parameter rename cannot depend on precision or on the fire option; the full regression suite was not rerun because the change is a rename with no behaviour change, and the IBSEB deck is compared byte for byte above.

Upstream

Source/ImmersedBoundarySEB is upstream as of erf-model#3960 and has the identical signature, so the same rename is wanted there; otherwise the warning returns with the next development sync. Happy to open that PR next.

🤖 Generated with Claude Code