Only write UTF-8 BOM where needed

[mgabeler-lee-6rs]

Describe the feature or problem you’d like to solve

Not all CLI text editors handle UTF-8 BOMs "nicely" -- haivng it visible and needing to navigate over it, and be sure not to leave it in the middle of an edited file (such as a merge message) is not great for users.

Proposed solution

The code that writes the BOM has a comment indicating that it exists just as a hint/workaround for notepad.exe on Windows:

cli/pkg/surveyext/editor_manual.go

Lines 43 to 44 in a9d397b

	// The reason why we do this is because notepad.exe on Windows determines the
	// encoding of an "empty" text file by the locale, for example, GBK in China,

If that comment is accurate, then the BOM writing could simply be limited to only running when runtime.GOOS == "Windows"?

If this is safe, it will make message editing a better, and less error-prone, experience for users on platforms that don't need it.

[mislav]

If that comment is accurate, then the BOM writing could simply be limited to only running when runtime.GOOS == "Windows"?

Thanks for raising this! I'd be up for only writing the BOM when GOOS = windows. I can't imagine that other editors/OS require the BOM to be present.

[mgabeler-lee-6rs]

🤔 Another way to make this more broadly safe would be to only write the BOM if we're in a non-utf8 locale. I could imagine other editors on any platform having similar behavior to notepad (given an empty file, assume the system locale), but if you're in a utf-8 locale, then that should be moot