This obsoletes `shlex`. The quoting turned out unnecessary, and the
splitting supported a lot more than we need. It also forced
unncessary allocations: The splitting doesn't add any characters and
keeps UTF-8 intact, so returning slices from the input is perfectly
possible.
Though this particular implementation will only come to use in the
future, as `CmdLine` still requires that the slices are cloned.
Still, the custom implementation performs about 3x faster.
There is no benefit from making all operations on binary data read
in small, async steps directly from the file.
Since we know file sizes beforehand, it's less complex to read
the entire data blob into memory once, then operate on that
synchronously.