Adds a wrt-init kind of on the model of git init for starting a stub
directory with wrt.json, archives/, publish/, and templates/.

Also starts to tweak POD a bit.

I went and read:

https://pause.perl.org/pause/authenquery?ACTION=pause_namingmodules

Realized I should probably not name this something with an inscrutable
top-level namespace.  WRT is still inscrutable, I guess, but at least
it's now under App.

- Renames all modules, fixes tests (such as they are)
- Zaps old t/validate/ stuff

This hadn't been in use for a while anyhow.

- create an index.html with new entries
- create an /all/index.html with a list of all entries
- create a /feed containing an atom feed of the most recent month
- a paragraph on requirements in the installation section
- copyright date through 2017

This will probably eventually bite me, since it could overlap with other
entry display logic, but it doesn't seem to break anything at the moment.

This is a fairly p1k3-specific thing.

- adds include_process() in WRT::Markup
- moves eval_perl() into WRT::Markup.  The way object instances are
  composed (and instance methods called) here still feels pretty janky,
  and this does nothing to help the situation.
- changes all existing $root_dir instances to $entry_dir; adds a $root_dir
  which represents the root of the archive (top-level folder containing
  the wrt.json file)

I'd be worried about the $root_dir change if I thought this module had a
single other user in the wild, but I strongly doubt that it does.  If I
am wrong and you are that user, I apologize sincerely in advance.

The include feature still has fairly unsettled semantics, but it gets the
whole thing much closer to a usable site generator.

Now uses:

  wrt display
  wrt render-all
  wrt fcgi

It's shorter and the letters are all right there on the keyboard.

There's nothing quite like finding a stupid bug in your stupid code you
wrote when you were like 20.

Anyway this turns out to be a nearly-pointless micro-optimization, but in
principle I think it's not wrong.

Stupid.

This reverts commit 4767c623c7.

Something is broken - only rendering a single year.

It appears to be a _lot_ faster to check "is a regular file" (knowable
just from stat, presumably) than a heuristic for "is ASCII or UTF-8", and
it's completely reasonable to treat regular files matching entry patterns
as markup blobs.

This whole sub is gross.

Really HTML::Entities should be exposed to templates and used there,
sorta like SparkLib\Template offers $h().