Port OreSpawn 4.0.16 to Forge 1.17.1 - #290

Summary

Backports the tested OreSpawn 4.0.10 through 4.0.16 campaign from Forge 1.18.2 to Forge 1.17.1, finishing at 4.0.16.117011.

Fixes included

• evaluates Stable Layers rock ranges against actual world Y while retaining shifted formation identity;
• binds the active dynamic biome registry before profile baking, combines exact and dictionary weights before climate fallback, and stores managed-ore filters as stable biome keys;
• aligns public sampler height and direct quart-biome attribution with chunk geology;
• adds immutable ore-dimension biome include/exclude ID and dictionary builder/getter contracts using the existing provider JSON fields;
• accepts canonical namespaced and legacy unnamespaced geome IDs;
• converts exposed one-layer Snow at top + 1 while preserving buried/authored Snow and Ice;
• leaves benchmark reporting and shutdown to GameTestServer, while ordinary dedicated benchmark servers still stop automatically.

Forge 1.17 adaptations

• retains Forge 37.1.1, pack format 7, Java 16 bytecode, reobfuscation and all six access-transformer contracts;
• uses Forge 1.17 BiomeDictionary plus stable ResourceKey<Biome> values because later biome-tag APIs are unavailable;
• shares a direct quart-biome lookup through ChunkBiomeContainer / BiomeManager.NoiseBiomeSource;
• keeps Gradle on exact Java 17 while compiling with the pinned Java 16 toolchain;
• adds a real cold Forge bootstrap job with the verified zero-patch result—no Mavenizer or Java 25 compatibility layer is introduced.

API major 1, schemas 4/6/5, registry IDs, existing provider field meanings, profile filenames, migration behavior, existing chunks and the [4.0.6,5.0.0) compatibility range remain unchanged.

Validation

• 128 tests across 33 suites, zero failures/errors/skips;
• 28 JSON/mcmeta files parsed; 16 locales with 357 matching keys; 21-document parity;
• full offline clean/check/build/Javadocs/release audit/checksum lifecycle;
• ForgeGradle 7 genEclipseRuns and production-classpath verification in both the primary and clean exact-commit worktrees;
• empty-cache online Forge bootstrap followed by same-cache offline replay with no source patch;
• exact reobfuscated packaged client plus fresh/reload server, including editor round trips, surface, natural-host, dynamic-biome, weather, fluid and attribution gates;
• Java 16/Forge 37/pack 7/six-AT artifact audit with no fixture, oracle, test, world or agent-note leakage;
• three matched 81-chunk runs: Sky/Average 14.898, Cyano 12.162 and vanilla 15.069 ms/chunk; normalized Sky change versus the accepted 4.0.9 control is +7.48%, below the 10% investigation threshold;
• clean runtime logs and no surviving benchmark process.

The exact candidate 3b73f396ffc15da66848162e38d37c990968297f was manually accepted before publication. This PR does not tag, publish, or directly update the MMD branch.