Closes #85
Closes #252

Alright, this took me a week to implement.

1. Name::is_valid must check chars() instead of bytes(), because the spec talks about "Any character with a Unicode value in the ranges [...]". E.g. the name "hello💝" encoded as bytes happens to include a banned control character but is a valid name.
2. user_name_to_file_name now...
   • takes a &Name, which guarantees some properties relevant for file name generation already
   • ensures a file name does not end with a period or space, as that's disallowed on Windows
   • ensures a file name is not using a Windows-reserved stem like "con", but unlike the official spec example does not replace more than necessary ("con.txt" must be turned into "_con.txt" but "hello.con.txt" is fine)
   • ensures a file name is unique among a case-insensitive set of existing file names. Or it panics after 99 attempts.
   • tries to not do something stupid if some form of usize over- or underflow occurs, e.g. by one of name, prefix or suffix being pathologically long. I'm not too confident about it though... I should probably fuzz it?
   • returns a PathBuf, because that's what's needed down the code chain. This allows me to simplify Layer::new and streamline its callers.
3. LayerSet and Layer are now responsible for providing user_name_to_file_name with a set of lowercased strings of existing file names for clash checking.
4. Layer::new now expects callers to do the Name instantiation work up front.

I think clashes are very unlikely, so this does feel a bit excessive... but I also can't rule them out completely. At least, I only try one strategy to resolve clashes instead of the two in the UFO spec example.

There is a discussion at unified-font-object/ufo-spec#164 about a different and better file naming scheme, but so far without a definitive conclusion.

I need to look more closely at Font::save() to see how to implement erroring out when a file should be unique but the filesystem thinks differently. Should Glyph::save_with_options() error out if the destination exists already?

Also, random observation: do we need to change or regenerate file names when removing/renaming glyphs? Doesn't seem like it.