Add end-of-run summary output to `logira run`

[melonattacker]

Motivation

logira run -- <command> records a run, but users currently need to run follow-up commands such as logira view last or logira explain last to understand what happened.

For the main use case — running Claude/Codex with strong permissions in a local environment — users should immediately know:

• Did anything suspicious happen?
• How many exec/file/net events were recorded?
• Were any detections triggered?
• How risky does the run look?
• Which command should I run next to inspect details?

The end-of-run experience should make logira useful after a single run.

Goal

Display a concise summary automatically when logira run finishes.

Example:

[logira] run 20260507-143022-claude finished in 2m34s, exit=0
  events:      47 exec, 123 file, 18 net
  detections:  2 high, 1 medium, 0 low
  risk:        HIGH

  top detections:
    HIGH   F021   read aws credentials/config: ~/.aws/credentials
    HIGH   E001   possible curl|sh execution pattern
    MED    F132   write shell startup file

  next:
    logira view 20260507-143022-claude
    logira explain 20260507-143022-claude --show-related

Scope

Implement end-of-run summary output for logira run.

The summary should include:

• run ID

• run duration

• audited command exit code

• event counts by type

  • exec
  • file
  • net

• detection counts by severity

  • info
  • low
  • medium
  • high

• simple run risk level

• top detections

• next-command hints

Risk level

For v0.2.0, risk scoring should stay simple:

HIGH   = one or more high-severity detections
MEDIUM = one or more medium-severity detections and no high detections
LOW    = one or more low/info detections and no medium/high detections
NONE   = no detections

This does not need to be configurable in v0.2.0.

CLI design

Add a summary flag:

logira run --summary auto -- claude
logira run --summary off -- claude
logira run --summary detections -- claude

Suggested behavior:

• auto: default human-readable summary
• off: suppress summary
• detections: show only detection/risk summary

The summary should be written to stderr so that the audited command's stdout is not polluted.

Non-goals

This issue should not implement:

• report export
• Markdown/HTML output
• auto-capture
• watch mode
• cloud upload
• UI
• configurable risk scoring

Acceptance criteria

• logira run -- <command> prints a concise summary after the command exits.

• The summary includes event counts.

• The summary includes detection counts by severity.

• The summary includes a simple risk level.

• The summary includes useful next-command hints.

• --summary off suppresses the summary.

• The audited command's exit code is preserved.

• The audited command's stdout is not mixed with summary output.

• Tests cover:

  • no detections
  • low/medium/high detections
  • non-zero audited command exit
  • summary disabled

Notes

This is the only must-have issue for v0.2.0.

The v0.2.0 goal is to make a single logira run immediately understandable.