fix(security): redact CLI credential output

[saagpatel]

Summary

• redact credential-shaped values at the shared CLI output boundary
• replace legacy LGTM annotations with current CodeQL suppression comments at five sinks already guarded by credential rejection or count-only contracts
• add regression coverage for named credentials, bearer authorization, provider tokens, and ordinary security-count output

Security scope

Targets CodeQL alerts #438, #440, #444, #445, #446, and #447. The generic CLI sink is hardened; the other five alerts are guarded false positives documented at the exact sink.

Verification

• 38 focused tests passed
• Ruff passed on changed Python files
• git diff --check passed
• Full suite: 3326 passed, 8 skipped, 27 failed only because optional serve/semantic dependencies were not installed in the isolated worktree; exact-head CI is required before merge.

[chatgpt-codex-connector]

💡 Codex Review

Here are some automated review suggestions for this pull request.

Reviewed commit: f4af8d0270

ℹ️ About Codex in GitHub

Your team has set up Codex to review pull requests in this repo. Reviews are triggered when you

• Open a pull request for review
• Mark a draft as ready
• Comment "@codex review".

If Codex has suggestions, it will comment; otherwise it will react with 👍.

Codex can also answer questions or update the PR. Try commenting "@codex address that feedback".