viewshed(): validate scalar params (max_distance >= 0, finite observer/target elev) before dispatch

[brendancol]

Describe the bug
viewshed() does not validate its scalar parameters before dispatching to a backend. Passing max_distance=-1 falls through into an empty spatial window and raises a cryptic error from deep inside the dask path:

ValueError: zero-size array to reduction operation minimum which has no identity

The same loose validation applies to observer_elev and target_elev: NaN, inf, or non-numeric values are not caught up front and produce confusing downstream errors instead of a clear message.

Expected behavior
The public viewshed() entry point should validate its scalar parameters before any backend dispatch:

• observer_elev and target_elev must be finite numeric scalars (reject NaN, inf, and non-numeric types) with a ValueError that names the offending parameter.
• max_distance must be a finite numeric scalar and >= 0. A negative value such as -1 should raise a clear message like max_distance must be >= 0, got -1.

Additional context
Validation lives before dispatch, so a single guard covers all four backends (numpy, cupy, dask+numpy, dask+cupy). This matches the existing scalar-validation style in sibling modules such as curvature.py (the cell-size finite check).