The latter should be useful when fully implementing all required
baseline capabilities of doing a correct checkout

…one (#301)

This was probably a left-over from times where there was a static
requirement on the chunks processing. Maybe… .

For now they are unused but that should change when doing collision
checks.

Let's not try to imitate git here but instead solve the problem our
way while learning from bugs git ran into in the process.

This will be particularly interesting when dealing with parallel
checkouts, something we need to implement for sure.

On the bright side, there seems to be opportunity for a performance
boost due to less iops during clone.

It takes more than 60 seconds to run this test, which is odd but for
now it seems ok to just dial down the numbers a little.