Remove ambiguous downloads from the datafiles.

[sh41]

This pull request resolves a issue where the data generation script could produce ambiguous or duplicate release names, causing the plugin to install an unintended JDK flavor.

The Problem

Previously, when multiple builds of the same Java version existed with different feature sets (e.g., jcef, fastdebug, lite), the script would strip these features and generate the same canonical name for all of them. This made it impossible for users to reliably select a specific variant and could lead to the installation of an incorrect or non-functional JDK.

The Solution

1. Deduplication: The script now uses a new jq query that:

   • Canonicalizes Keys: It generates a unique key for each release based on its vendor, type, and a predefined set of allowed features (musl, javafx, lite, etc.).
   • Intelligently Selects a Primary Entry: It groups all releases that share the same canonical key and selects the one with the most minimal feature set as the single, authoritative entry. This eliminates duplicates without discarding valid distributions.

2. Tests I've moved tests into their own script and added a check that all the release names are unique.

I also spotted some upstream data issues which I've opened tickets for: https://github.com/joschi/java-metadata/issues/78 https://github.com/joschi/java-metadata/issues/80 https://github.com/joschi/java-metadata/issues/81

Fixes: https://github.com/halcyon/asdf-java/issues/255