feat(jira): add Jira Cloud context-source adapter

[mattsenicksigma]

Summary

Closes #174.

Adds a Jira Cloud context-source adapter that ingests issues filtered by project key and label allowlist, converts them to staged JSON, and runs the jira_ingest memory-agent skill to produce durable wiki knowledge.

• New adapter at packages/cli/src/context/ingest/adapters/jira/ with fetch, chunk, detect, client, ADF-to-Markdown converter, and JiraSourceAdapter class
• driver: jira connection schema added to driver-schemas.ts and registered in local-adapters.ts and public-ingest.ts
• jira_ingest skill at packages/cli/src/skills/jira_ingest/SKILL.md for the memory agent
• since date filter — optional YYYY-MM-DD field that narrows JQL to issues updated on or after that date; no per-run cap
• ADF (Atlassian Document Format) description field recursively converted to Markdown before staging
• Pagination via GET /rest/api/3/search/jql with nextPageToken (old /rest/api/3/search endpoint removed by Atlassian)
• One work unit per project key (jira-<projectkey>)
• 20 new tests (ADF converter, fetch, chunk)

ktx.yaml example

connections:
  jira-docs:
    driver: jira
    base_url: https://yourorg.atlassian.net
    email: you@yourorg.com
    api_token_ref: env:JIRA_API_TOKEN
    projects:
      - DOCS
      - ENG
    labels:
      - kb
      - decision
      - policy
    since: "2024-01-01"   # optional — omit to fetch all matching issues

Test plan

• pnpm --filter @kaelio/ktx run test — all Jira adapter tests pass
• pnpm --filter @kaelio/ktx run type-check — clean
• pnpm run dead-code — clean
• End-to-end: ktx ingest <connection-id> with a real Jira token fetches issues, stages JSON, and the memory agent writes wiki candidates

🤖 Generated with Claude Code

[vercel]

@mattsenicksigma is attempting to deploy a commit to the Kaelio Team on Vercel.

A member of the Team first needs to authorize it.

[mattsenicksigma]

A more human comment

Pretty much the same flow as #319 -> just reading in some text and adding to the wiki generation.

novelty here is parsing out the text from the issues. An intermediary conversion to markdown is levereaged as it seemed like a reasonable way to standardize things.

Obviously this an #319 are more straightforward, #316 is the more novel/interesting of the three adapter pr's