docs(blog): address reviewer feedback on logarithmic quantities

Pawel Trella and Chip Hogg reviewed the article in the comments. Changes:

- The opening example now pairs a watt with a volt on the same `6.0 * dB`, and
  "Gains carry their domain" closes the loop with the domain-typed fix.
- `level + level` is ill-formed because the combination is underdetermined by
  correlation and phase, not merely because it needs linearizing.
- New "Digital audio full scale" section: `dBov` (full scale, also the origin a
  DAW peak meter uses) and `dBFS (AES17)` 3.01 dB below it, with peak versus RMS
  placed on the quantity spec rather than the unit. No bare `dBFS` is defined.
- The nholthaus survey now describes 3.x, where `dBW + dBW` is deleted, and keeps
  only what still differs (hardcoded `10 * log10`, power-only decibels, the
  dimension-based level/gain split, no neper or frequency intervals).
- `level + gain` states its rule: matching classification, and a kind that is
  either the level's own or dimensionless.
- Open Question 7 (chaining transfer functions in the log domain) and Open
  Question 8 (whether a gain carries its domain), both raised by the review.

Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>

Author: mpusz