`search_data` should warn when spatial parameters are used with a "no-spatial" collection

[danielfromearth]

Related to #1305 (fix in #1309) and #515.

When the bounding_box parameter (probably for polygon and point parameters too) is passed to search_data() for a collection with no spatial metadata (e.g., TEMPO_IRR_L1, version V04), the search silently returns an empty list. There is no warning or indication that the spatial parameter is the reason for zero results, and the silent [] leads users to believe the data simply don't exist.

This is the inverse of #515, where spatial filters silently match everything for global collections:

Scenario	Collection Type	Spatial filter effect	User experience
#515	Global (every granule covers full Earth)	Matches everything	"Why didn't my bbox narrow the results?"
This issue	No-spatial (no spatial metadata)	Matches nothing	"Why is there no data?"

Both could benefit from a collection-level check of spatial metadata to inform users when their spatial parameters won't behave as intended.

Reproducing

import earthaccess

# Returns 45 results — no spatial filter
results = earthaccess.search_data(
    short_name="TEMPO_IRR_L1",
    version="V04",
    temporal=("2023-08-01", "2025-09-20"),
)
print(len(results))  # 45

# Returns 0 results — bounding_box silently filters everything out
results = earthaccess.search_data(
    short_name="TEMPO_IRR_L1",
    version="V04",
    temporal=("2023-08-01", "2025-09-20"),
    bounding_box=(-180, -90, 180, 90),
)
print(len(results))  # 0 — no warning

Possible Approach

Before or during granule search, earthaccess could check the parent collection's UMM-C for spatial coverage type. If it indicates no spatial coverage and spatial search parameters are present, earthaccess could emit a warnings.warn(), optionally drop the spatial parameters, or raise an informative error. Perhaps this could also be handled partly in python-cmr. A similar check could potentially address #515 as well.

Context

We encountered this at NASA ASDC with TEMPO irradiance data, a solar measurement with no Earth spatial footprint. The GES DISC team independently hit the same kind of problem with gnssro_cosmic1_jpl_l1b (#1305, though that issue focuses on the __repr__ exception). This likely affects any collection without spatial metadata — solar irradiance, calibration, and other non-Earth-pointing products.

Environment

• earthaccess: 0.15.1
• Python: 3.11.13