feat: consolidate governance implementation and skill updates

[leefowlercu]

Summary

This PR establishes a consistent distribution and governance model for the 20 Terraform and Packer Agent Skills while
expanding Terraform provider-development guidance.

• consolidates all Skills under plugins/terraform/skills/<skill> and plugins/packer/skills/<skill>
• publishes the terraform and packer product Plugin Bundles through both the Claude Code and Codex marketplaces
• removes six legacy Plugin identifiers without compatibility aliases and documents the replacement installation paths
• adds and expands Terraform provider configuration, resources, actions, scaffolding, documentation, testing, framework
  migration, ephemeral resource, and Terraform Search/import guidance
• adds lifecycle metadata, a 20-Skill catalog, explicit Skill ownership, and a six-model supported model matrix
• adds contribution, security, support, issue-intake, and pull-request guidance
• removes the Tessl-based evaluation workflows and the repository-local Terraform Policy evaluation suite
• validates marketplace parity, Plugin metadata, Skill metadata, catalog entries, ownership coverage, and Markdown links

User and Contributor Impact

Individual installation with npx skills remains the primary installation method. Every supported path is listed in
SKILLS.md.

Claude Code and Codex both publish exactly the terraform and packer product Plugin Bundles. The following legacy
Plugin identifiers are removed without aliases:

• terraform-code-generation
• terraform-module-generation
• terraform-provider-development
• terraform-policy-code
• packer-builders
• packer-hcp

The README documents how to replace those identifiers and migrate individual Skill paths to
plugins/<product>/skills/<skill-name>.

Contributions remain limited to internal HashiCorp contributors. CONTRIBUTING.md, SECURITY.md, and SUPPORT.md
describe the contribution boundary, sensitive-reporting guidance, and repository support boundaries.

Supported Models

SUPPORTED_MODELS.md publishes the governed latest and N-1 baselines for the GPT, Claude Opus, and Claude Sonnet model
families. The GPT baselines are GPT-5.6 Terra as latest and GPT-5.5 as N-1.

Evaluation and Validation

Tessl-based evaluation workflows and public evaluation assets are removed. This PR does not introduce replacement
evaluation CI.

Validation completed for this branch includes:

• structural and JSON validation for two marketplaces, two product Plugin Bundles, and 20 Skills
• Claude Code and Codex installation of both product Plugin Bundles in isolated environments
• individual npx skills installation for all 20 Skills
• Claude Code and Codex Plugin schema validation
• Skill schema validation for all 20 Skills
• local and external Markdown link checks
• scratch compilation and tests for the new Go provider examples
• shellcheck, bash -n, and git diff --check