chore: drop redundant full-forge naming

[MrCoder]

Summary

Project is pure Forge across all variants (lite/full/diagramly), so the full-forge label — a leftover from the Connect→Forge migration — is redundant. This cleans up dead code and stale docs across the full variant pipeline.

• tests/e2e-tests/config/apps.ts — removed dead Connect-era zenuml-full@stg profile (pointed at zenuml-stg.atlassian.net); renamed zenuml-full-forge@stg → zenuml-full@stg. Local pnpm test:full:stg now correctly hits the live Forge full app on full-stg.atlassian.net.
• .github/workflows/build-test-deploy.yml — E2E: Full now uses app: zenuml-full@stg.
• .github/workflows/staging-deploy.yml — dropped dead full-forge ternary; install-step condition flipped from full-forge → full. Behavior change: the staging-full job will now run forge install --upgrade on full-stg.atlassian.net after each deploy. Previously this never fired (no caller passed full-forge), so the staging Full app silently stayed pinned to whatever was last installed manually.
• .claude/skills/{validate-branch,ship-branch,land-pr,babysit-pr}/SKILL.md — "4 variants" → "3 variants (lite, full, diagramly)".
• operations/full-forge-app.md → operations/full-app.md; title updated.
• Deleted full-forge-pipeline.md (described a dual Connect+Forge pipeline that no longer exists).
• CHANGELOG.md left as-is (historical record of the actual release label at that date).

Affects all variants in CI plumbing, but the runtime behavior change is full only.

Test plan

• pnpm vitest run --dir src → 160/160 passed
• python3 -c "import yaml; yaml.safe_load(...)" on both touched workflows → valid
• grep -rn "full-forge" excluding CHANGELOG → no matches
• CI green on this PR (staging deploys + lite E2E once marked Ready)
• After merge, watch the first staging-full run on main confirm the new install-upgrade step succeeds (continue-on-error, but should be green)