[mmacferrin]

ivert.vdatum_lookup built its name table from transformez.definitions.Datums, which transformez 1.0 removed. The ImportError was caught silently, so every current install has run on IVERT's short built-in table alone, and some of its codes were wrong:

Name	Mapped to	Problem
mlw	EPSG 1091	A vertical datum code, not a CRS; transformez rejects it
mhw	EPSG 5868	transformez has no operation registered for it
mtl	EPSG 5713	CGVD28 height (Canada), not Mean Tide Level
lwd_igld85	EPSG 9000	ITRF2014, a geographic CRS

-V mlw and -V mhw failed partway through a validation.

This PR:

• Passes tidal datums to transformez as its own IDs. mllw, mlw, msl, mhw and mhhw resolve to vdatum:mllw etc., leaving transformez to decide how each surface is applied. Geoid and ellipsoid names still resolve to EPSG codes.
• Checks every datum against transformez before a run starts. -V accepts a name, an EPSG code or a transformez ID, and stops at once, with transformez's reason, on one it cannot transform.
• Drops names transformez 1.0 cannot transform: ngvd29, igld85, mtl, lwd_igld85 and navd88 height (usft). transformez's VDatum support does list the MTL and DTL surfaces but has no reference to reach them through yet; a comment in vdatum_lookup.py says to add mtl and dtl once it does.
• Carries a transformez ID through validation. dem_geom.split_srs_string() splits a compound CRS whose vertical part pyproj cannot read (EPSG:4326+vdatum:mllw); transform_points(), validate_dem and plot_photon_clouds_v2.py use it. validate_dem()'s dem_vertical_datum now resolves names itself.
• Rewrites vdatum_lookup's API: resolve_vdatum(), plus new check_vdatum(), describe_vdatum() and list_vdatums(), which ivert validate --list-vdatums now uses. The descriptions come from transformez.

Datums that worked before give the same results. Comparing main against this branch:

• Transforming 500 points near Chesapeake Bay to navd88, mllw, msl and mhhw gives identical heights.
• ivert validate on a CRM Vol 6 tile with -V navd88 and with -V mllw gives identical results files and summary statistics.

The only difference is the manifest, which records vdatum:mllw where it recorded EPSG:5866.

CHANGELOG entries under Fixed and Changed, including a Breaking entry for the dropped names.

---

🔍 Docs preview: https://ivert--132.org.readthedocs.build/en/132/