Match the scan transcript to the format strings that print it

[kridaydave]

Problem

The scan transcript in the README is not what the tool prints. The runtime and container rows are indented four spaces with the column padding starting at position four, which is how a hand-aligned table looks. The code prints two spaces of indent with the pad starting at position two.

Evidence

The code prints:

• src/cli.rs:609 — println!(" {:<20} {} (pinned)", r.name, r.version);
• src/cli.rs:620 — println!(" {:<20} {} → {}", c.name, c.image, c.version);

Rendered, node produces 38 characters beginning node 20.5.0 (pinned). The README line 29 was 41 characters beginning node 20.5.0 (pinned).

I reproduced both format strings and compared them to the README byte for byte. Line 29 and line 30 were each 41 characters against a real 38, and line 33 was 46 against a real 38. The → arrow was already correct, and the 41-character divider matched all sixteen in src/cli.rs, so only the three rows drifted.

Change

Three lines, indent and padding corrected. All three now match the format strings exactly. The mount transcript below them is untouched and was already correct.

Not shipped

• action.yml:36-39 declares a registry-path input. scripts/run.sh:11 assigns it to REGISTRY_PATH and never reads it again. Deleting an input from a published GitHub Action changes its public interface, so it needs the owner's call rather than a drive-by.
• README.md:135 cites corpus counts ("23 bottlenecks, 21 sources") from a different repo, so this repo cannot verify them.