feat(du): add du disk usage filter with compact output

[TheAgentWire]

Summary

Adds rtk du — a compact filter for du (disk usage) output. Closes #284.

84 invocations in the last 30 days (per issue reporter) — this is a frequently used command that benefits from token-optimized output.

What it does

• Strips whitespace padding from sizes
• Sorts by size descending (largest first) — immediately surfaces what is eating disk
• Strips common directory prefix from paths for shorter output
• Shows total as Σ at bottom — clearly distinguishable from directory entries
• Graceful fallback on filter failure (raw du output)
• -h flag passthrough — disable_help_flag prevents clap from swallowing -h (human-readable), since du -h is the most common usage pattern. Help is still accessible via --help.

Examples

rtk du -sh /path/to/project/src/*

Before (raw du):

 32K    src/aws_cmd.rs
 60K    src/cargo_cmd.rs
 40K    src/cc_economics.rs
 12K    src/ccusage.rs
8.0K    src/config.rs
 28K    src/container.rs
104K    src/discover
 68K    src/main.rs
 60K    src/git.rs

After (rtk du):

104K    discover
68K     main.rs
60K     cargo_cmd.rs
60K     git.rs
40K     cc_economics.rs
32K     aws_cmd.rs
28K     container.rs
12K     ccusage.rs
8.0K    config.rs

Sorted largest-first, common src/ prefix stripped, padding normalized.

rtk du -h -d 1 /path/to/project

1.2G    target
1.9M    .git
1.2M    src
316K    docs
84K     scripts
36K     .github
20K     hooks
Σ 1.2G

Root directory detected as total, shown with Σ marker. Immediately obvious that target/ dominates.

rtk du -sh /path/to/workspace/*

3.1G    project-alpha
1.5G    project-beta
1.2G    project-gamma
212M    project-delta
48M     project-epsilon

Quick disk audit across all projects, sorted by size.

Token savings

≥60% verified in tests. Biggest wins on recursive output with long absolute paths — common prefix stripping + sort + padding removal compound significantly.

Files changed

File	Change
src/du_cmd.rs	New filter module (run, filter, sort, prefix strip, total detection, 10 tests)
src/main.rs	Register mod, Commands enum variant, match arm
src/discover/rules.rs	Hook rewrite pattern + rule for du → rtk du
tests/fixtures/du_*.txt	4 real fixture files for test reference

Testing

• 10 unit tests: single entry, multi entry, sorting, empty input, raw blocks, common prefix stripping, parent path total detection, size parsing, token savings ≥60%
• Full suite: 718 tests pass, 0 failures
• Manual testing: verified on macOS with real directories
• Pre-commit gate: cargo fmt --all --check && cargo clippy --all-targets && cargo test ✅

Checklist

• Unit tests added for all filter logic
• Token savings ≥60% verified
• Edge cases covered (empty, single, raw blocks, parent total)
• cargo fmt && cargo clippy && cargo test passes
• Manual test: rtk du inspected and verified
• DCO sign-off included
• No personal/sensitive data in fixtures or tests

[pszymkowiak]

Hey @TheAgentWire, thanks for this work!

However, since PRs #349, #351, #386 landed, RTK now includes a built-in TOML filter for du (see src/filters/du.toml). Tested savings: 96.8% on du -h . with large directory trees.

The TOML filter covers the core use case declaratively in ~6 lines. Closing this PR as the functionality is already shipped.

If you'd like to contribute more advanced filtering (e.g., sorting, aggregation), feel free to open a new PR building on top of the TOML filter. Thanks again!