Fix: tighten email regex to reduce false positives in email detection

[DivInstance]

Summary

This PR reduces false positives in email detection used across Slash by replacing the very loose \S+@\S+ with a stricter, centralized regex.

Changes

• Add EMAIL_RE in api/extract.py
  • Pattern: \b[a-zA-Z0-9._%+-]+@[A-Za-z0-9.-]+\.[A-Za-z]{2,}\b
• Update extract.mail()

Rationale

The previous pattern matched tokens with trailing punctuation or random strings containing @, causing misclassification (e.g., treating the input as an email when it’s not). The new pattern aligns with common email formats and reduces noise.

Testing

• Positive: [email protected], [email protected] → detected as emails
• Negative: hello@world, username@, @domain.com, not-an-email. → not detected

Impact

Improves accuracy for flows that depend on [extract.just.mail()]

Checklist

• Centralized email regex
• Updated both email extraction methods
• Manual smoke tests for positives and negatives

Resolves #2