improve buffer URI format

[rchl]

Per discussion in #2360, change the URI format for unsaved files so that the part after the scheme is parsed as a path and not as an authority (netloc in python).

Shouldn't really affect anything (neither fix the Pyright issue) but seems like a more correct thing to do.

(Also workarounds the issue with urlparse failing to parse the URI correctly when path starts with numbers - add View- prefix).

[jwortmann]

Might be better to change the format to buffer:view-<id> or something like that rather than adding a workaround for urllib.

This prefix is also just a workaround for the urllib bug, it's just a different one.

Actually it is the buffer ID, not the view ID. I can push my test changes from yesterday so that we could compare; I think I'd slightly prefer the other (explicit) workaround, and it might improve a few other places in the code too. But I'm not sure yet.

[jfcherng]

Not Buffer- prefix? 🤪

[jwortmann]

Two reasons why I probably prefer the other approach:

• The underlying bug is that urllib can't parse the buffer:123 format correctly, and not that the format would be invalid somehow. So I think the fix/workaround should be for the parsing, and not to modify the URI format.
• We should keep in mind what might be a potential format that could natively be used by ST in the future (even if it might never be added): Implement view.uri() sublimehq/sublime_text#3742
  I assume that buffer:123 should be more likely than buffer:Buffer-123.