spec: expand radar precipitation scope to single-radar products

[franchg]

Summary

• Previously the radar precipitation spec explicitly excluded single-radar datasets and only accepted multi-radar composites. This PR expands the scope to also admit single-radar products, provided they meet the existing §3.2 spatial requirements (≤1 km resolution, and valid sensing area supporting at least one 256×256 crop).
• Motivation: discussion with AEMET on including the Zaragoza regional radar dataset (~1 km, 480×480, fits a 256×256 crop within the sensing disk). The composite-only rule was a proxy for "enough pixels for a decent receptive field" — §3.2 already enforces that directly, so the blanket exclusion was redundant and overly restrictive.

Changes

• §1 Introduction: "2D gridded radar … datasets" instead of "composite datasets".
• §2 Scope: both multi-radar composites and single-radar products are in scope, subject to §3.2.
• §3.2 Spatial Requirements: reword resolution clause to "2D gridded radar data"; clarify that the 256×256 sensing-area rule applies equally to composites and single-radar products.
• CHANGELOG: add [Unreleased] > Changed entry.

Follow-ups (not in this PR)

• check_spatial_requirements currently validates array dimensions, not the valid-data sensing disk. A single-radar product with heavy shadowing could pass by dims alone. Enforcing the disk-level requirement needs a mask/NaN convention that isn't yet in the spec — worth a separate discussion.
• Leif noted that §4.4 should render the allowed standard_name values explicitly in the spec docs. Not included here.

Test plan

• uv run mkdocs build --strict produces a clean spec render with the updated §1 / §2 / §3.2 text.
• Existing tests still pass (no check logic changed in this PR).

[leifdenby]

Looks great to me - thanks for making these changes!