feat: implement Chorus.ai MCP server with full API coverage

[Ompragash]

Introduce a complete Model Context Protocol server for the Chorus.ai (ZoomInfo) conversation intelligence platform. The server exposes 39 tools, 6 URI-based resources, and 6 workflow prompts that enable LLMs to interact with sales call data, transcripts, scorecards, and analytics through natural language.

Architecture decisions:

• JSON:API normalization layer in the API client auto-flattens Chorus responses (data/attributes/id) into plain objects, keeping all 12 tool domain files decoupled from the wire format.
• Annotation presets (READ_ONLY, CREATE, DELETE) defined as typed constants eliminate repetition across 39 tool registrations while preserving full MCP spec compliance.
• Dual transport support: stdio (default) for CLI/desktop clients, Streamable HTTP for networked deployments. Transport selected via TRANSPORT env var at startup.

Tools by domain (39 total):
Conversations (5), Users (3), Teams (3), Scorecards (4), Playlists (3), Moments (4 incl. create/delete), Emails (2), Engagements (2), Reports (3), Saved Searches (3), Video Conferences (4 incl. upload/delete), Integrations/Session (3)

Workflow prompts:
chorus_call_analysis, chorus_deal_risk_assessment, chorus_competitive_intelligence, chorus_meeting_summary, chorus_rep_performance_review, chorus_customer_feedback_synthesis

Infrastructure:

• Shared services: Axios-based API client with auth injection, HTTP error-to-actionable-message mapper, markdown/JSON formatters with CHARACTER_LIMIT truncation
• Zod schemas with .strict() validation on all tool inputs
• Pagination via page[size] and cursor on every list endpoint
• Jest test suite: 125 tests across 9 suites (unit, integration) covering services, all tool domains, and tool registration integrity
• CI: GitHub Actions matrix testing on Node 20/22/24
• Release: tag-triggered npm publish with provenance to @opensourceops/chorus-mcp