`spatial` arrays pick the first *truthy* Location, not the first *usable* one

[cody-seibert-gsa]

Code repo: GSA/datagov-harvester

Problem

When a v3.0 spatial value is an array of Location objects, _unwrap_location() (general_utils.py:1250-1251) selects the first element that is truthy, not the first that is actually usable (i.e. has geometry/bbox/centroid):

if isinstance(input_value, list):
    input_value = next((item for item in input_value if item), None)

A non-empty dict with only prefLabel is truthy, so it wins over a later element that actually carries geometry:

Input	Result
[{bbox: POLYGON(…)}, {prefLabel: "Alaska"}]	✅ Polygon
[{prefLabel: "Alaska"}, {bbox: POLYGON(…)}]	❌ NULL

Order in the source feed silently determines whether a dataset gets a translated_spatial at all.

Separately (lower priority, same function): even when selection works, only one of N array elements is ever kept, so a multi-region spatial silently drops every region but the first. spatial_shape is a geo_shape OpenSearch field, so combining into a MultiPolygon/GeometryCollection is possible if that's wanted — that part is a product call, not filed as a fix here.

Suggested fix

In _unwrap_location, iterate the list and return the first element that actually unwraps to something usable (has a resolvable geometry/bbox/centroid, or is itself a bare GeoJSON dict), rather than the first truthy one.

Acceptance criteria

• [{no usable geometry}, {usable geometry}] resolves using the second element
• [{usable geometry}, {no usable geometry}] is unchanged (still resolves using the first)
• Existing test_translate_spatial_location_array_uses_first still passes
• Unit test added for the reversed-order case — there is currently none

Context

Found in #6038 (SPIKE-3), finding 3.