app-modeling: validate implicit runtime semantics

[willdavsmith]

⚠️ Radius application graph diff unavailable because this skills repository does not contain a deployable app.bicep.

Summary

Require app-modeling to validate implicit runtime semantics before emitting Bicep: application parser coercion and unset behavior, artifact-level recipe output mappings, container-image platform defaults, and authentication/bootstrap readiness.

Reason for change

A SQLPad + Azure SQL evaluation compiled cleanly while producing an undeployable model. It treated a read-only SQL Server port schema property as available even though the selected Azure recipe maps only host, inherited an incompatible amd64+arm64 image-build default for a Dockerfile that executes target binaries, and omitted the access bootstrap needed for the primary API path. Compilation alone did not detect these defects.

Fixes #

How to test

1. Run app-modeling against sqlpad/sqlpad@ab1f0c03269f0178b9449d34505ce3462271f340 with the Azure SQL profile.
2. Confirm generation opens the exact Azure provider recipe pack, uses only mapped outputs, uses the proven Azure SQL port literal, selects a Dockerfile-compatible platform, preserves source parser semantics, and configures a usable access path.
3. Compile the generated file and deploy it through the SQLPad Azure sample E2E workflow with a Radius.Security/secrets recipe registered by the Environment.

File change summary

File	Summary of change
skills/app-modeling/SKILL.md	Adds artifact-level recipe evidence, runtime parser, platform-default, and access-readiness requirements to generation and validation.
skills/app-modeling/references/runtime-contract.md	Expands the runtime ledger and consistency checks for exact recipe mappings, string coercion, provider-backed ports, image platforms, and bootstrap access.
skills/app-modeling/references/bicep-structure-rules.md	Prevents incompatible implicit image platforms and references to schema-only read-only outputs.