v0.5.15

Added

• TUI output now shows confidence score after the entity label,
  dimmed: └── PHONE_NUMBER (σ 0.9) (σ = score, 1 decimal).
• Phone findings now emit US_PHONE (NANP, country code 1) or
  INTL_PHONE (everything else) as distinct entity types. Disable
  independently via [entities]:

  [entities]
  INTL_PHONE = false  # only worry about US phones

  PHONE_NUMBER in [entities] still disables the whole recognizer
  for a bigger hammer.

Fixed

• CC false positive on 6.349667550340612: Luhn + entropy accidentally
  passed for this float's digits. Added the same embedded-in-identifier
  check used by phone: reject if preceded/followed by alphanumeric, .,
  or _.
• Email false positive on supercharge/redis-github-action@1.8.1:
  GitHub Actions version pins look like emails. Now reject when the TLD
  has no letters (per RFC 1035, real TLDs always do).
• Phone false positive on 1234567890 / 1234567891: plain-digit
  phones require a context keyword on the same line (phone, tel,
  mobile, contact, etc.) and must pass NANP structural validation
  (area code and exchange first digit in 2-9). Rejects sequential test
  IDs and NPI numbers.
• Phone false positive on 3.214.229.114: phone with 4 dotted groups
  and a 1-3 digit final segment is shaped like IPv4, not a phone.
• Phone false positive on 760292141147 in AWS ARNs: long plain-digit
  runs now require phone-context keywords on the line.
• IPv4 false positive on OIDs like "id": "1.2.543.1.34.1.34.134":
  reject IPv4 matches that are embedded in a longer dotted-numeric
  identifier (digit+dot immediately before, or dot+digit immediately
  after the match).

Changed

• Post-emit entity-type filter on strict findings, so setting e.g.
  INTL_PHONE = false filters individual findings without disabling
  the PhoneRecognizer as a whole.

Full Changelog: v0.5.14...v0.5.15