[Lane 2 · kotlin-core] GroupNormalization — implement GroupNormalization.forward #1381

Sub-issue of #1379 ([Feature]: GroupNormalization forward pass (groupNorm {} DSL currently crashes)).

Lane: 2 · Kotlin core
Skill needed: Kotlin. You do not need to know SKaiNET internals beyond the tensor ops used in LayerNormalization.kt.
Size: s (a few hours)
Blocked by: nothing

What to do

1. Open skainet-lang/skainet-lang-core/src/commonMain/kotlin/sk/ainet/lang/nn/normalization/GroupNormalization.kt and LayerNormalization.kt side by side.
2. Replace the four TODO() helpers: reshape [N, C, H, W] → [N, G, C/G*H*W], take mean/variance over the last dim, normalize with eps, reshape back.
3. Replace the VoidOpsTensor gamma/beta placeholders with real parameters the way LayerNormalization creates them, and apply them per channel.
4. Remove @NotImplemented from the class.
5. Add a test next to the existing normalization tests using the Lane-1 values (or values you compute with PyTorch yourself).
6. Run ./gradlew :skainet-lang:skainet-lang-core:jvmTest checkUnmarkedStubs, then regenerate the matrix with ./gradlew :skainet-compile:skainet-compile-hlo:jvmTest --tests '*HloCoverageDumpTest' :skainet-compile:skainet-compile-dag:jvmTest --tests '*BackwardCoverageDumpTest' generateDocs and commit the changed docs/modules/ROOT/pages/reference/ files.

Acceptance

• Forward matches PyTorch within 1e-5 for all Lane-1 cases
• checkUnmarkedStubs passes with the TODOs gone
• Matrix NN-layers table shows ✅ for GroupNormalization
• Result reported back on the parent issue

Notes

If something here turns out to be wrong or much bigger than the size label, comment on this issue or open a specific bug instead of blocking the parent.

Coverage matrix: docs/modules/ROOT/pages/reference/ops-status-matrix.adoc · How lanes work: Issue taxonomy