See #43

Takes a package name and a dest path. This way we can more eassily change the
layout and it will error out on package name typos or missing packages.

Add a scope so we don't have to care about naming conflicts
and set the verison to 0.1.0 since we don't have a stable API yet.

We already do all this in createVideoElement

We no logner use that library and the "qr-scanner" altternative turned out to
be better, so remove it completely.

We now queue it and in the callback check if the scan is still active,
and go from there. This means we redraw all the time and detect mostly
in the background.

This results in worse perf with the old library since that isn't async, but
we can always revert if needed.

This allows the detection to be async and pasue execution between capturing
the image, analyzing it and then finally drawing it.

With both being async it's hard to follow what would happen if both run at the same
time, so just prevent it and serialize start/stopping instead.

so 'this' is the same

This lets the user decide what to do (like scrolling)