fix(skills): compact Jira implementation reads

[jhonDoe15]

Summary

• require explicit JSON summary mode whenever TWG skills recommend
  --agent-fields
• distinguish identity/status @compact reads from implementation-bearing Jira
  reads
• add a standard-library helper that requests bounded Jira fields and converts
  ADF descriptions to compact Markdown JSON
• project formal Jira issue links and report unsupported ADF nodes/marks instead
  of silently claiming complete evidence
• cover field projection, ADF headings/lists/tables/marks, unsupported content,
  batch payload shapes, and command construction

Refs #18

Public Safety Checklist

• No tokens, credentials, cookies, auth files, or .env files.
• No private Jira, Confluence, Bitbucket, infrastructure, pipeline, or
  customer URLs.
• No customer data, proprietary content, or internal runbook detail.
• Generated docs/skills/plugins came from the release/public profile.
• Plugin or release metadata points to public TWG CLI docs and public CDN
  URLs.

Verification

node --check skills/twg-jira/scripts/jira-workitem-implementation.mjs
node --test skills/twg-jira/test/jira-workitem-implementation.test.mjs

Four tests pass. A local authorized smoke test reduced a detailed Jira
workitem from approximately 53.5 KB of native JSON to 6.4 KB of projected JSON
while preserving all requirement section headings and reporting no unsupported
ADF features; no private content is included in this PR.

The repository's existing public-safety workflow currently fails on main
because skills/twg-bench-lite/scripts/benchmark-lite/runner.mjs contains the
literal auth.conf, which the workflow's pattern flags. This PR introduces no
new safety-pattern matches.

[atlassian-cla-bot]

Hooray! All contributors have signed the CLA.