fs.globSync returns incorrect results with non-UTF8 encoded file names.

For example... on Linux, where file names are just byte sequences...

const { writeFileSync, existsSync, globSync } = require('node:fs');

const path = Buffer.from([0xe9]);
writeFileSync(path, 'hello');
existsSync(path);  // true

console.log(globSync('[^a-z]')); // [ '�' ] non-utf8 bytes in the file name end up replaced.

The globSync, et. al APIs likely need an encoding option that instructs it to return the filenames as Buffer objects.

Start with the fs.globSync entry point and reproduce the Linux example using a filename containing the byte 0xe9. Trace how glob results are decoded and check the related glob APIs for consistent behavior. Done means non-UTF8 filenames can be returned without replacement characters, with the proposed Buffer-oriented encoding option covered by tests.