feat(aidd-context): map the project ecosystem in memory

[blafourcade]

Problem

Project memory describes the codebase but not the ecosystem around it. No file says which external tools the project runs on, which one is authoritative for what, or how they hand work to each other. A skill that needs to know where a User Story actually lives, in a file or in a tracker, has to ask again every session.

Two concrete symptoms in the current templates:

• plugins/aidd-context/skills/02-project-memory/assets/templates/memory/core/vcs.md:9 carries Ticketing: <jira, linear, none>. That is a backlog fact living in the VCS file, and the backlog memory added by feat(aidd-pm): orchestrate a product backlog #559 now owns the same thing. Two homes, one of them wrong.
• plugins/aidd-context/skills/02-project-memory/assets/templates/memory/README.md exists, but it maps the memory/ folder itself: how it loads, which files it holds. It does not map the ecosystem.

Nothing else covers it: core/architecture.md stops at the tech stack, deployment/deployment.md at the pipeline, core/codebase-map.md at the source tree.

Scope

• Add an ecosystem memory file: the external tools the project uses, what each one is authoritative for, and how they interact. One row per tool, conventions only.
• Move Ticketing out of core/vcs.md into the backlog memory, which already has a Supports table for exactly this.
• Audit every other template for a fact that belongs to another concern, and list what was found.
• Wire the new file through references/memory-destinations.md, references/capability-signals.md, and the scaffolded tree in references/structure.md.

Acceptance criteria

• Reading the memory alone answers: which tool holds the backlog, and does an artifact live in a file or in that tool
• Ticketing appears in exactly one memory file
• Every template is checked for a fact owned by another concern, and each finding is either moved or justified
• The ecosystem file is produced by the scan from repo facts, like every other concern, not hand-written
• A project with no external tool gets no empty file

Prior art in this repo

• assets/templates/memory/backlog/backlog.md (feat(aidd-pm): orchestrate a product backlog #559) — its Supports table already models "support, authority for, role". The ecosystem file should reuse that shape rather than invent one.
• references/capability-signals.md — how a concern is detected from a concrete repo fact; the ecosystem file needs its own signals.
• references/memory-destinations.md — the table that decides where a template lands.
• assets/templates/memory/README.md — the existing README, kept as the folder map, not extended into an ecosystem map.

Out of scope

• Reading live data from an external tool. Memory records durable conventions, never work items, live values, or technical ids. backlog/backlog.md already states this and the new file follows it.
• MCP server configuration. Which tools are reachable is a runtime concern, not a memory one.