Delete USEEIO-recreation flags without ongoing consumers

[briantobin-99]

cc:
Closes:

What changed? Why?

Deletes the USEEIO-recreation flags that have no remaining consumer, plus their gated code:

• use_ghg_national_2023_m2 — the m2 FBS path, the output-weighted NAICS→BEA mapping, and the CH4-GWP-27.9 workbook override. The electricity NAICS retargeting hook in the standard mapping path is untouched.
• useeio_margins — the USEEIOR margin filter set and inflation_helpers_useeio.py (the USEEIOR CPI table). get_rho_inflation_ratio stays (the publish Excel Rho panel uses it).
• ceda_margins — the legacy CEDA filter set and derive_2017_margins_ceda_usa / derive_phi_ceda_usa.
• use_useeio_schema — read by no code; its only effect was the m2 validator interlock.

Kept: scale_a_matrix_with_useeio_method, use_E_data_year_for_x_in_B, and deflate_x_to_detail_io_year_for_B remain in use as diagnostic probes by the electricity-disaggregation work (year_alignment_bly_e.py, the electricity configs, and the mixed-units handling inside the D/N_new_inflated block), so they are out of scope here.

Configs deleted: useeio_phoebe_23.yaml and v03_waterfall_useeio_g1_schema_ghg.yaml — both set the deleted margins/m2 flags; the USEEIO-track waterfall sheets stay pinned in the registry, and the CEDA-track waterfall configs are untouched.

Margins comparison scripts: compare_margin_approaches and compare_phi_to_reference (studies of the settled margins decision) are deleted, with an archived output run stored externally as the record. The two compare_sef_* scripts are kept on pinned inputs: --phoebe-sef-csv is now required (a phoebe SEF published before this merges serves as the pin), and compare_sef_zenodo_useeio_code defaults to 2025_usa_cornerstone_v0_3.

Testing

Unit suite, black, ruff, and mypy green. None of the deleted paths are active in the v0.3 config; Snapshot integration suite (test_integration run) passes on the stack including this change.

[WesIngwersen]

Margins-project handling (the sign-off question): compare_margin_approaches and compare_phi_to_reference (June studies of the settled margins decision) are deleted. the two recent compare_sef_* scripts are kept and converted to pinned inputs: --phoebe-sef-csv becomes required (publish a phoebe SEF once before merge and pin it), and compare_sef_zenodo_useeio_code defaults to 2025_usa_cornerstone_v0_3. If live phoebe publishing is still needed for Q3 margins work, hold this PR.

I'm making sure we have a record of the margins analysis output linked to a git commit which we can store internally so we have the analysis...i will open up a PR with updates i made to be able to run it off main; store the outputs on GDrive, then I am fine with deleting the margins analysis code and configs stated in this PR, as long as the team is happy with the record of output being enough data/visuals to explain margin changes from useeio and ceda.

[WesIngwersen]

FYI @briantobin-99 please review and pull this one in before merging this. Otherwise I am good with this one.