Expose typed worktree admission separately from capacity warnings

[chubes4]

Problem

workspace worktree add reports a single warning-oriented capacity status while downstream callers need to distinguish advisory worktree-count pressure from blocking byte or inode refusal. Today count-only pressure allows creation, while --force only overrides true byte/inode refusal, but compact machine-facing output omits explicit admission and override fields. Consumers can infer the wrong policy from status: warning.

Acceptance

• Compact structured output includes creation_allowed, force_override_required, and typed trigger reasons with advisory or blocking severity.
• Count-only warning reports allowed/no force.
• Byte/inode refusal reports blocked/force required.
• Human output distinguishes advisory pressure from refusal in one stable line.
• Tests cover count-only, byte-only, inode-only, and combined cases.

AI assistance

OpenAI GPT-5.6 Sol through OpenCode audited DMC admission code, live capacity output, and existing issues and drafted this report. Chris Huber reviewed and remains responsible.