output: braille summary visuals for scan Summary (TTY only)

[buttonmonkey]

What

Two small, dependency-free visual touches to the scan Summary, both TTY-only:

1. Findings heat map — the "Most affected files" list gains a per-file braille bar: length = finding count (normalized to the busiest file), colour = worst severity in that file (reusing ui::severity_code).
2. Severity mix bar — a single stacked braille bar under the ● N critical ● N high … line, segments sized by each severity's share and coloured by severity.

What it looks like

Colour doesn't render in GitHub code blocks — run it in a terminal to see it. Below, bar length = count and bar colour = severity (critical = bold red, high = red, medium = yellow, low = green).

A real multi-file scan — heat map (per file) + severity mix bar (overall blend):

Summary
  ● 2 critical   ● 7 high   ● 1 medium   ● 1 low
  ⣿⣿⣿⣿⣿⣿⣿⣿⣿⣿⣿⣿⣿⣿⣿⣿⣿⣿⣿⣿⣿⣿⣿

     7  Command Injection
     2  Unsafe Deserialization
     1  Dynamic Import
     1  Open Redirect

Most affected files
  ⣿⣿⣿⣿⣿⣿⣿⣿⣿⣿     4  ./app/api/handlers.py
  ⣿⣿⣿⣿⣿⣿⣿⡇⠀⠀     3  ./auth/session.py
  ⣿⣿⣿⣿⣿⠀⠀⠀⠀⠀     2  ./db/query.py
  ⣿⣿⡇⠀⠀⠀⠀⠀⠀⠀     1  ./utils/format.py
  ⣿⣿⡇⠀⠀⠀⠀⠀⠀⠀     1  ./web/redirect.js

A balanced severity mix — the stacked bar reads the blend at a glance:

Summary
  ● 5 critical   ● 5 high   ● 5 medium   ● 5 low
  ⣿⣿⣿⣿⣿⣿⣿⣿⣿⣿⣿⣿⣿⣿⣿⣿⣿⣿⣿⣿⣿⣿⣿⣿

     5  Command Injection
     5  Dynamic Import
     5  Open Redirect
     5  Unsafe Deserialization

The mix bar is normalized to the total, so it shows the blend of severities rather than the volume.

Pipe-safety (the important guarantee)

Both visuals are gated behind ui::color_enabled() (TTY and NO_COLOR unset). When stdout is piped or NO_COLOR is set, the output falls back to the original plain list, byte-for-byte unchanged, and the JSON/CSV paths are untouched. Verified by diffing pre- vs post-change output:

• sighthound sample | cat — identical
• NO_COLOR=1 sighthound sample — identical
• sighthound sample -o json / -o csv — identical

No new dependencies

Braille is packed by hand (U+2800 base), in the spirit of the dependency-free ui.rs. Cargo.toml / Cargo.lock are untouched. Bars are padded to a fixed 10 glyphs so the count column stays aligned; deterministic ordering (BTreeMap + existing sorts) is preserved.

Tests

Adds unit tests for the pure helpers in output.rs:

• braille_bar — the char-count == width alignment invariant (across widths/fracs), fill math, out-of-range clamping, and partial-cell behaviour.
• severity_rank — ordering including the unknown case.

Full suite green; no existing tests changed.

Try it

cargo build --release
./target/release/sighthound <a directory with a few findings>

Any tree with findings across several files / severities will show both visuals in a colour terminal; piping it (| cat) shows the unchanged plain output.

Scope: a single file, src/scanner/output.rs.

[asadeddin]

@buttonmonkey

Beautiful UI/UX contribution.

A couple of quick questions:
1- Any concern over accessibility for color blindness?
2- Most affected files, does it carry over the color scheme or will it always be red?

[buttonmonkey]

1. Potentially, the bars are mapped to your classification colors: critical, high, medium, low. Not sure of how to mitigate that as I don't know how to create a reliable pattern fill or some equivalent that would work in this setting. If I were to find that pattern, it would require changing your color values to match the pattern fills.
2. The bars should equate to the classification types. Only the same color if they are all of the same type. (ex. all critical)