[GLUTEN-CORE] ResourceUtil fails to load JARs from paths with special characters (URL-encoded)

## Description

`ResourceUtil.getResources()` fails when Gluten is loaded from a JAR file located in a directory with special characters in its path (such as `@`, spaces, or other characters that get percent-encoded in URLs).

## Root Cause

`ClassLoader.getResources()` returns `URL` objects where special characters are percent-encoded (e.g., `@` → `%40`). `ResourceUtil` passes these URL-encoded paths directly to `new File(url.getPath())`, which expects filesystem paths, not URL-encoded strings. This causes `File.exists()` to return `false`, resulting in a `GlutenException`.

## Example

When JARs are cached in a path like:
```
~/.cache/coursier/v1/https/user%40domain.pkgs.visualstudio.com/project/_packaging/feed/maven/v1/org/apache/gluten/gluten-package/...
```

The `%40` in the path causes `ResourceUtil` to look for a non-existent directory/file.

## Fix

Use `URL.toURI()` and `new File(URI)` which properly decode percent-encoded characters, while also correctly preserving `+` characters (unlike `URLDecoder.decode()` which treats `+` as space).

PR: #11672

Start at ResourceUtil.getResources(), which is identified as the failing entry point. Reproduce loading Gluten from a JAR path containing spaces or other percent-encoded characters, then verify that resources load successfully without a GlutenException and that literal '+' characters remain intact.

Written by the indexing model from the issue text.