Bump icy_sixel to 0.6.0 (consolidates Dependabot #207)

[l1a]

Consolidates Dependabot #207 onto a gated branch so the release hygiene it bypasses — version bump, NOTES.md entry, man regen — is actually done. Follows the #167/v0.6.3, #184/v0.6.16, #188/v0.6.19 and #199/v0.8.1 pattern.

retch-cli → 0.9.5. retch-sysinfo unchanged at 0.1.56.

What changed and why

icy_sixel 0.5.1 → 0.6.0. Applied as Dependabot's exact patch, so this carries precisely the change #207's green CI validated; Cargo.lock was then diff-verified to be that entry plus our own retch-cli bump, with no transitive movement. Worth stating, because quantette — icy_sixel's palette quantiser — moved in v0.8.1, and a move here would have changed emitted colours rather than just structure. It did not.

This is not a lockfile-only roll. A caret range on a 0.x spec will not admit 0.6, so the Cargo.toml spec widened "0.5" → "0.6" — the same shape as v0.6.4's base64 0.22 → 0.23, and the reason that bump was deliberately excluded from the v0.6.3 consolidation. icy_sixel is graphics-feature-only, so just check's --features graphics pass (v0.6.5) and the CI graphics-feature job (v0.6.7) are the only legs that compile it at all — a green default matrix proves nothing here.

A 0.x minor is semver-breaking by convention, but the break does not reach retch — established from the source, not from the fact that it compiled. The only two calls retch makes (SixelImage::try_from_rgba and .encode(), one call site in src/logo.rs::print_sixel_rgba) live in sixel_image.rs, which together with encoder.rs is byte-identical between the two versions. The entire 0.6.0 change is decoder-side: a new stateful SixelDecoder whose colour registers persist across images (purely additive in lib.rs's exports), plus a MAX_PIXELS guard against decode-time memory exhaustion. retch never decodes sixel — it only ever emits it.

The 164-line docs/retch.1 diff is not from this bump

just man removed 164 .Bl/.El lines. Cause: mandown was upgraded to 1.1.1 after v0.9.4 shipped, and it no longer emits those around list items.

The new page is the correct one. .Bl/.El are mdoc macros, undefined in man(7): groff reports macro 'Bl' not defined and macro 'El' not defined on the committed page and zero warnings on the regenerated one, while the rendered output is byte-identical (409 lines, same sha256). Nothing a reader sees changed; two warnings went away.

Proved pre-existing rather than assumed, by regenerating at the committed version 0.9.4 first and confirming the diff was exactly 164 deletions, all .Bl/.El, zero other lines either way. Without that control, a 164-line man-page diff arriving beside a dependency bump looks like the bump's doing.

This is the v0.6.2/v0.6.16 flip-flop on a new axis, and it flips the same way: a contributor on an older mandown regenerates, re-adds the macros, and just pr step 4 fails with a large diff that reads as "my change broke the man page." The wiki's Development-Setup.md now pins mandown ≥ 1.1.1 and names the symptom.

Test plan

• just pr green end to end — fmt, clippy --workspace and --features graphics, 254 tests (87 cli lib + 15 integration + 152 sysinfo), man regen idempotent at retch 0.9.5, Cargo.lock committed, audit advisory-only (pre-existing allowed paste RUSTSEC-2024-0436)
• Cargo.lock diff-verified against Dependabot's deps(deps): bump icy_sixel from 0.5.1 to 0.6.0 in the cargo-dependencies group #207: exactly the icy_sixel entry, no transitive drift
• encoder.rs and sixel_image.rs sha256-compared across 0.5.1 and 0.6.0 — identical
• Isolated encode of a fixed 64×48 RGBA gradient through try_from_rgba + encode is identical on both versions (9885 bytes, FNV-1a fe96948691ac471a), with the probe's own lockfile read to confirm it really resolved 0.6.0
• End-to-end: real --features graphics binary in a 138-col PTY with TERM=xterm-sixel emits an identical sixel payload on both versions (ca8a8d19…, 10203 bytes, 228 colour registers, well-formed DCS introducer and ST terminator), with each build's lockfile verified at build time
• Man page: groff warnings 2 → 0, rendered output byte-identical (409 lines, same sha256)
• Man page delta proved pre-existing by regenerating at the committed version 0.9.4
• Wiki pushed before opening this PR — Development-Setup.md, 4476f52..48627da

Legitimate skips, each checked rather than assumed: README.md and docs/retch.md (no user-visible field, CLI flag or config key changed; both mention Sixel only as a capability, with no version), Configuration-and-Theming.md (no config key, theme or display flag changed) and Workspace-Architecture.md (no module layout or crate-boundary change). No wiki page names icy_sixel or pins a dependency version.