the asset system is reworked to support "dynamic" entries, where each
entry is a separate file on disk containing the latest generation's
headers+raw+gzip+zstd. when calling view.regenerate, it will look for
pages that had "export const regenerate" during generation, and render
those pages using the view system, but then store the results as assets
instead of sending as a response.
pages configured as regenerable are also bundled as views, using the
non-aliasing key "page:${page.id}". this cannot alias because file
paths may not contain a colon.
"server side rendering" is a misleading term since it implies there is a
server. that isn't neccecarily the case here, since it supports running
in the browser. I think "clover engine" is cute, short for "clover html
rendering engine". Instead of "server side rendering", it's just rendering.
This commit makes things a lot more concise, such as `ssr.ssrAsync`
being renamed to `render.async` to play nicely with namespaced imports.
`getCurrentRender` and `setCurrentRender` are just `current` and
`setCurrent`, and the addon interface has been redesigned to force
symbols with a wrapping helper.
the problems with the original implementation was mostly around error
handling. sources had to be tracked manually and provided to each
incremental output. the `hasArtifact` check was frequently forgotten.
this has been re-abstracted through `incr.work()`, which is given an
`io` object. all fs reads and module loads go through this interface,
which allows the sources to be properly tracked, even if it throws.
closes#12
