SDK: Fosberg dead fuel moisture grids (new grid family)

[amarcozzi]

Wrap the generated create_fosberg_fuel_moisture_grid endpoint. This is a new grid family (fuel_moisture) with no existing SDK surface — e.g. create_dead_fuel_moisture_grid_from_fosberg(...).

CreateFosbergFuelMoistureRequest is a grid → grid derivation (no external raster, no alignment; output inherits the topography grid's georeference):

• source_topography_grid_id (needs slope + aspect bands, degrees)
• source_irradiance_grid_id (a Leaflux irradiance grid with irradiance.surface.relative — see the Leaflux issue)
• dry_bulb_temp (°F, >= 10)
• relative_humidity (percent 0–100)
• time (HHMM, restricted 0800–1959)
• month (FuelMoistureMonth)
• elevation (RelativeElevation: below/near/above — a Fosberg correction category, NOT the topography elevation band)

Depends on #195. Consider ordering after the Leaflux issue since Fosberg consumes an irradiance grid. Add live-API test coverage.

[amarcozzi]

Heads-up from the API side: the blocker referenced here (silvxlabs/FastFuels-API-v2#547) is fixed and deployed to api-v2-prod — silvxlabs/FastFuels-API-v2#548.

Along with unblocking the irradiance grid, that PR adds a synchronous 422 same-lattice contract on the two-source grid-derivation endpoints, which the SDK wrapper should be aware of:

Fosberg (create_fosberg_fuel_moisture_grid) — source_topography_grid_id and source_irradiance_grid_id must share the same horizontal lattice: equivalent CRS, trailing (y, x) shape, and affine transform (the output inherits the topography grid's georeference). The irradiance grid may be 3D (canopy+surface) — only its y/x lattice is compared. Mismatch → 422 with resample guidance.

LeafLux (upstream) — the optional terrain grid must likewise sit on the LAD grid's lattice, or 422.

Both endpoints also now scope every source-grid lookup to the domain, so a source grid from another domain → 404.

Practical implications for the SDK work here:

• Live-API test coverage should build the topography and irradiance grids on the same lattice (same domain, resolution, and extent), otherwise create_fosberg_... will 422 before dispatch.
• Worth surfacing the 422 clearly and documenting the same-lattice requirement in the wrapper's docstring; the OpenAPI field descriptions for both source_*_grid_id fields were updated to state it and how to resample.