Guard unsupported reaction slice admission

[woksin]

Added

• Add regression coverage for rejecting unsupported Automation and Translate slices without returning a usable semantic compilation (Automation and Translation slice types are not admitted by the Cratis ESM planner Stage#79).

[woksin]

Parent source review found an incomplete contract change. This PR is now draft, not merge-ready.

SemanticModelBinder.BindSlice admits Automation/Translate after this change, but SemanticModelCanonicalJson.SliceKind and SemanticModelRead.ParseSliceKind still support only StateChange/StateView. Adding enum values does not preserve occurrence/effect semantics or implement reactions, and the existing syntax renderer is not a safe substitute for a semantic renderer.

Regression specs are being run against the newly admitted bare slice kinds before correcting the admission boundary. The earlier passing existing suite had no coverage of this new behavior. No dependency or release action is being taken. Independent Sol review attempts exhausted provider quota without a verdict; these are parent source findings, not a completed independent approval.

[woksin]

Corrective commit b850bbe is pushed, preserving the original commit in branch history. The net diff against current main is now only two regression-spec files (eight facts), not new semantic slice kinds.

Red/green evidence:

• Before correction: 8 executed, 6 passed, 2 failed on the required unsupported-syntax diagnostic. Binding caught the serializer contract error; no uncaught exception or successful compilation was observed.
• After correction: 2,288/2,288 language specs and 22/22 canonical checks passed on net10.0, including all eight new facts.
• Strict Debug and Release Rebuilds passed with warnings treated as errors.
• All four corrective candidate hashes were unchanged during verification. Both production files now exactly match upstream main.

The PR remains draft because independent review has not returned a verdict. This does not implement Automation/Translation semantics or resolve Stage #79. Future work needs actual portable occurrence/effect contracts and a semantic renderer; simply routing to the legacy syntax renderer is not sufficient.