GitHub: PR system should enforce ASCII and whitespace conventions

[StephanTLavavej]

clang-format enforcement (#88 resolved by #132) is amazing, but I anticipate the need for more enforcement. Specifically:

• We have a convention of requiring all sources to be ASCII. It's easy to accidentally introduce non-ASCII characters (e.g. Standardese often contains smart quotes and long dashes).
• We avoid tabs in most files. It's currently possible for tabs to sneak into // clang-format off [...] // clang-format on regions.
• We have whitespace conventions equivalent to VSCode's

"files.insertFinalNewline": true,
"files.trimFinalNewlines": true,
"files.trimTrailingWhitespace": true,

clang-format doesn't enforce the final newline conventions. It does trim trailing whitespace, but only in source files; other text files aren't covered. (And again, trailing whitespace is a concern in regions where we've disabled clang-format.)

• We're using CRLF line endings. Nothing is preventing files with LF line endings from being introduced.
• (optional) We have a convention of wrapping to 120 columns. clang-format does this in source files only, but it's imperfect (we've encountered situations where it'll happily exceed 120).

I've written a fairly simple file scanner that detects many of these issues; I could clean it up for production use. Of course, it's a C++ program that has to be compiled.