feat: Add `sentry review` command for AI-powered code review

[betegon]

Summary

Add a sentry review command that uses Sentry's Seer AI to review code for bugs, logic errors, and issues that could cause production failures.

Why Before the PR?

The key insight: catch problems before you push, not after.

When you create a PR:

• CI pipelines spin up (costs time and resources)
• Teammates get notified
• Failed checks create noise and delay

With sentry review, you can run a quick AI-powered check locally and fix issues before creating the PR. Ship cleaner code on the first push.

Command Design

# Review current branch changes (vs main)
sentry review

# Review specific PR
sentry review --pr 123

# Review a file or directory
sentry review src/handlers/

# Review staged changes only
sentry review --staged

What It Catches

• Logic errors - Bugs and flawed control flow
• Error-prone patterns - Code likely to cause production failures
• Performance issues - Regressions or inefficient code
• Security concerns - Potential vulnerabilities

Key Differentiator

Unlike generic linters, Seer combines:

1. Error context - Historical production errors from Sentry
2. Performance data - Real issues your app has experienced
3. Code analysis - Static examination of changes

This surfaces issues that will actually cause production problems, not pedantic style nits.

Technical Details

• Use Seer API endpoint for code review
• Integrate with GitHub PR context when available
• Output actionable suggestions with file:line references
• Support both interactive and CI modes

References

• Product page: https://sentry.io/product/seer/ai-code-review/
• Similar to existing Seer commands (sentry issue explain, sentry issue plan)