refactor(render): colors in colors.rs, escape codes and sanitize in printer.rs

[o2sh]

Follow-up to #1888. On main, color and escape-code logic was spread across render:

• Two places wrote colored text: style::render for the info lines, and logo.rs, which wrote its own escape codes and skipped sanitize.
• printer.rs combined get_ascii_colors and TextColors by hand, with no_bold passed alongside.
• The language chips picked their own colors in fields/languages.rs.

Design

render/
  printer.rs   picks the output, puts the ascii art or image beside the info, writes it
  info.rs      the info: title, "Key: value" lines, palette row      (was lines.rs)
  fields/      the value lines of each field
  format.rs    numbers and dates
  ascii.rs     ascii art template -> lines                           (was logo.rs)
  image/       sixel, kitty, iterm
  style.rs     Line, Segment, Style
  colors.rs    Colors: what each Style looks like (color and bold)

Style names everything that can be colored, as the command line names it: the six parts of --text-colors (Title, Tilde, Underline, Key, Separator, Value), the {n} parts of the ascii art (AsciiArt(n), for --ascii-colors), the language chips and the languages bar (LanguageChip, LanguageBar), and the palette row (Palette(n)). It holds no colors.

• Only colors.rs decides how each Style looks: its color, from the language, --ascii-language, --ascii-colors, --text-colors and --true-color, and whether it is bold (--no-bold). The rest of render only says which Style a piece of text has.
• Only printer.rs writes escape codes (apart from the image protocols), so it sanitizes everything it prints, the ascii art included.
• ARCHITECTURE.md describes this layout and the two rules.

Options

InfoOptions and RenderOptions stay where they are on main, in data/mod.rs and render/mod.rs. This branch briefly moved them into options.rs (802fda3), but that made a loop: data and render imported options, which imports cli, which imports data and render types. da7dfc8 reverts it.

Behavior

The visible output is the same as main. Two differences in the bytes written:

• The ascii art no longer writes an empty escape pair (\e[…m\e[0m) before each {n} token.
• --ascii-input now goes through sanitize like all other text, so control characters in it (tabs, escape codes) show as �. None of the built-in ascii art contains control characters.

Testing

• cargo fmt, cargo clippy --all-targets: clean.
• cargo test: 155 unit tests, 12 integration tests and the doc test pass. New tests cover Colors::style (including the chip and palette styles), the chip colors, paint, sanitize and ascii_beside_info. The ascii tests now check Lines.
• Compared main and this branch on two repos (onefetch, and a repo whose manifest description contains an OSC escape sequence). I used 18 flag sets: default, --no-bold, -c, -t (full and partial), --true-color always/never, --ascii-language (alone and with --no-bold -c), --ascii-input (with and without {n} tokens), --no-art, --nerd-fonts, --no-color-palette, --no-title, --number-of-languages, -o json and -o yaml.
  • With escape codes stripped, the output is identical in all 36 runs.
  • The raw bytes are identical once main's empty escape pairs are removed (checked on 16 of the runs).
  • The escape sequence in the manifest description prints as �.
  • Churn and last change were disabled for the comparison, because they vary between two runs of the same binary.

🤖 Generated with Claude Code