Future development suggestions

[RCHowell]

While porting the initial generator work, I've realized this has become comically complicated. Here are some suggestions for future me or future others which have been on my mind.

• This is suffering from the inner-platform effect, reduce anodizer complexity
• We could drop the anodizer model/parser and all that since smithy is our frontend.
• Keep the filesystem and template abstractions as those are very useful.
• The options are likely not needed.
• Serde for smithy (and smithy-aws) builtins needs a solution.