Related to: #58576

Currently, if a benchmark file is encoded in utf16 (which is the default in windows), the benchstat tool will fail silently. This might be hard to debug to the end user.

I propose that instead of failing silently, the tool should return an error letting the user know their file is not properly encoded.

On windows, running this command: go test -bench . -count 10 > old.txt will return a utf16 encoded file by default.

It is a simple change, after this line, we just need to add something like:

if !utf8.Valid(line) {
	r.err = fmt.Errorf("%s: invalid encode, only utf8 encoded files are supported", r.result.fileName)
	return false
}

And then if the user were to run the benchstat tool, they would get that error instead of it failing silently.