serve: Giant refactor to change how we handle data when serving pages. * We need a distinction between source metadata and route metadata. In most cases they're the same, but in cases like taxonomy pages, route metadata contains more things that can't be in source metadata if we want to re-use cached pages. * Create a new `QualifiedPage` type which is a page with a specific route and route metadata. Pass this around in many places. * Instead of passing an URL around, use the route in the `QualifiedPage` to generate URLs. This is better since it removes the guess-work from trying to generate URLs for sub-pages. * Deep-copy app and page configurations before passing them around to things that could modify them, like data builders and such. * Exclude taxonomy pages from iterator data providers. * Properly nest iterator data providers for when the theme and user page sources are merged inside `site.pages`.

import os


def slashfix(path):
    if path is None:
        return None
    if isinstance(path, str):
        return path.replace('/', os.sep)
    fixed = []
    for p in path:
        fixed.append(p.replace('/', os.sep))
    return fixed