absence-scan: run main() on Windows, where the entrypoint guard never matched (#339)

The CLI exits 0 having done nothing on Windows, so the pre-push hook built on
it passes everything — including the bytes the scanner exists to refuse.

`import.meta.url === \`file://${process.argv[1]}\`` compares a URL against a
string built from a native path. On POSIX those coincide. On Windows
import.meta.url is file:///C:/... while process.argv[1] is C:\... with
backslashes, so the template produces file://C:\... and the comparison is never
true. main() is never called.

pathToFileURL is the same helper proxy/pipeline.mjs already uses to turn an
extension path into an import URL.

Measured on Windows at 8ddd4f0, before and after:

  node tools/absence-scan.mjs                     exit 0, no output   ->  exit 1 + usage
  node tools/absence-scan.mjs <file with a UUID>  exit 0, no output   ->  exit 2 + FINDING capture-uuid
  node tools/absence-scan.mjs <clean file>        exit 0, no output   ->  exit 0 + "absence-scan: clean"
  node --test test/absence-scan.test.mjs          11 pass / 8 fail    ->  19 pass / 0 fail

The eight failures are the existing CLI: and git-range: cases. They were
already encoding the correct contract; nothing on Windows was running it.

Author: deafsquad

tools/absence-scan.mjs

@@ -34,6 +34,7 @@
 import { readFileSync } from "node:fs";
 import { execFileSync } from "node:child_process";
 import { basename } from "node:path";
+import { pathToFileURL } from "node:url";
 
 // --- Allowlist ---------------------------------------------------------------
 //
@@ -467,7 +468,12 @@ function main(argv) {
   return 0;
 }
 
-if (import.meta.url === `file://${process.argv[1]}`) {
+// Entrypoint guard. pathToFileURL, NOT `file://${argv[1]}`: on Windows
+// import.meta.url is file:///C:/... while argv[1] is C:\... with backslashes,
+// so the template form never matches and main() never runs — the CLI becomes a
+// silent exit-0 no-op, and the pre-push hook that depends on it passes
+// everything. Same helper proxy/pipeline.mjs already uses for its loader.
+if (import.meta.url === pathToFileURL(process.argv[1]).href) {
   let code;
   try {
     code = main(process.argv);