Use CRS area-of-use bounds for coordinate clamping instead of proj4 string detection

[frewsxcv]

Context

When reprojecting raster extents to a target CRS, source coordinates may fall outside the target CRS's valid area of use. For example, EPSG:3857 (Web Mercator) is only defined for latitudes ±85.06° — near-polar vertices produce extreme Y values that stretch the projected extent.

Current workaround

We detect Mercator by checking if the target CRS's PROJ4 string contains +proj=merc, and clamp source latitudes to ±85.06° when it does:

let target_is_mercator = crs_definitions::from_code(self.target_crs.epsg_code)
    .map(|def| def.proj4.contains("+proj=merc"))
    .unwrap_or(false);
if target_is_mercator {
    const MERCATOR_LAT_LIMIT: f64 = 85.06;
    min.y = min.y.max(-MERCATOR_LAT_LIMIT);
    max.y = max.y.min(MERCATOR_LAT_LIMIT);
}

This only handles Mercator. Other projections with limited areas of use (e.g., UTM zones, Lambert conformal conic) are not clamped.

Ideal solution

Use a proper CRS area-of-use database to clamp coordinates for any target CRS, similar to PROJ's proj_get_area_of_use(). This would:

• Handle all projections with limited areas of use, not just Mercator
• Use the official EPSG-defined bounds rather than hardcoded constants
• Remove the fragile PROJ4 string parsing

Blockers

The crs-definitions crate we use only exposes code, proj4, and wkt — it does not include area-of-use/bounds information. Options:

1. Contribute area-of-use bounds to crs-definitions upstream
2. Use a different crate that includes bounds data
3. Build a lookup table from the EPSG database