feat(index): log resume progress when continuing an interrupted build

[raphaelsty]

Summary

When a resumable build resumes after an interruption (timeout, Ctrl-C, crash), it now prints a one-line summary of how much work is being skipped:

📋 Resuming interrupted build: 100 files already indexed, 200 remaining

The already-committed files are excluded from the todo set in build_resumable, so they are never re-parsed or re-embedded. Previously a resume looked identical to a fresh build from the outside, which made it hard to tell that prior work was being reused. This just surfaces the existing behavior.

Why

Came out of verifying the resumable-build guarantee end-to-end. I generated a ~12,000-unit synthetic repo (300 files) and interrupted indexing mid-build, both ways:

Interrupt	Committed at cut	After resume	Re-encoded on resume
SIGKILL (hard crash)	100 files / 4,100 docs	300 files / 12,300 docs	only the 200 remaining files
SIGINT (graceful Ctrl-C)	100 files / 4,100 docs	300 files / 12,300 docs	only the 200 remaining files

In both cases:

• The resume re-embedded only the not-yet-committed files — the 100 already-committed files were skipped entirely.
• The fully-committed chunk files were left byte-identical across the resume (only the trailing partial chunk was reopened to append new docs).
• The final index was identical to a clean full build — same file count (300) and document count (12,300), no duplication, no loss — and searched correctly.

So the build correctly resumes from the last checkpoint and never recomputes committed work.

Testing

• cargo build -p colgrep, cargo fmt/clippy -D warnings clean (pre-commit CI passed)
• Manual SIGKILL + SIGINT interrupt/resume cycles as above