Support approvals in local mcp

[nieblara]

Summary

Add 7 approval workflow tools to the MCP server, enabling AI agents to participate in the LaunchDarkly approval process for feature flag changes.

Changes

New tools (7)

• list-approval-requests — List all pending approval requests across the account
• list-flag-approval-requests — List approval requests for a specific flag/environment
• create-flag-approval-request — Submit flag changes for review (required when mandatory approvals are enabled)
• get-flag-approval-request — Get details of a specific approval request (status, reviewers, proposed changes)
• delete-flag-approval-request — Withdraw a pending approval request
• apply-flag-approval-request — Apply an approved request to make changes live
• review-flag-approval-request — Approve, decline, or comment on a pending request

Manual change

• schemas/mcp-enable-tools.yaml — Added overlay entries for 7 approval API endpoints

Generated (109 files, via speakeasy run)

• 7 func implementations (src/funcs/approvals*.ts)
• 7 MCP tool wrappers (src/mcp-server/tools/approvals*.ts)
• 7 operation models, 22 component models (request/response types for the approvals API)
• SDK client (src/sdk/approvals.ts), server registration, docs, examples

What I did

1. Added approval endpoint entries to schemas/mcp-enable-tools.yaml
2. Ran speakeasy run to regenerate — no manual edits to any generated file
3. Built and smoke tested via stdio transport: server starts, all 26 tools register (19 existing + 7 new)
4. Verified list-approval-requests returns real data