ci: fail on a BOM or text garbled by the wrong encoding

[anurag629]

What does this change?

Adds a CI step right after checkout that fails when a file starts with a UTF-8 BOM or has text garbled by the wrong encoding, and prints every file and line it found. It looks for the BOM and for a few byte sequences that only show up when UTF-8 was read as another encoding, like a dash that turned into three other characters. The patterns are written as byte escapes, so the workflow file doesn't match itself.

Both have reached PRs this month: four new files with a BOM in the first version of toolbelt #190, and 24 garbled lines in its lib/tools.ts after a rebase. Review caught them, but CI should.

How I tested this

• The step passes on the current base branch, so no file needs fixing first.
• In a scratch repo with a BOM file and a garbled line, it fails and names both lines. A file with real dashes, arrows and middle dots passes.
• Against the earlier commits of toolbelt #190 it flags the four BOM files on b7ef7fd, and the 24 garbled lines in lib/tools.ts plus the template's BOM on c7de664.
• The workflow still parses as YAML.