Resolve reliability issues with the feeders

[brainstorm]

At present, the feeders do not advance components on the tape reliably:

See issue markmaker/PushPullFeeder#9 and markmaker/PushPullFeeder#10 for context.

@PancakeLegend and myself have a good understanding about the parameters involved and how it's being problematic. Unfortunately, we haven't yet found a combination of (OpenSCAD) parameters for that model that result in flawless, continuous operation :_/

Also, please take a look at this post on the devs mailing list: https://groups.google.com/g/openpnp/c/XJbMe4WIDXk/m/1cketbxEBQAJ. In that post I detail an issue that prevents us from using one of the fully functional feeders because:

1. When setting up the ~4 movements the nozzle and feeder need to work, all offsets and positions look totally right.
2. When the machine is in operation, though, an extra offset is introduced, preventing the advance of the feeder (nothing wrong with the feeders, probably has to do with bottom vision calibration).

In other words:

For the PushPullfeeder, all the coordinates are handled relative to the two tape sprocket holes. The idea is that you can move the feeder around on the machine, remove it, mount it again, with only hand precision, and still get very precise pick locations.

For this to work, your tape transport needs to be very repeatable. If the tape transport is not moving the sprocket holes to very consistently the same location relative to the feeder each time you actuate the lever, all the locations will equally be affected. Now that I write this down, I see that is a potential for a chicken-egg situation 😬. It never happened in my (endless) testing, but I can see how it could. I guess you would have to manually actuate the feeder, so it is in its canonical transport position, then press Reset Statistics. This should bring you out of the chicken-egg situation.

To be a bit more specific: the "dog" thorn must reliably engage with the sprocket hole. Make sure your pull motion goes far enough to make sure of that, and add a bit of margin. Then the push motion must equally go far enough back, to push the dog all the way against the bar, with some extra tension (I think it is mentioned in the video). These together should make sure the transport is repeatable. I never had any problems that way, but that might have been sheer luck.

Can you help, @dmello-io? ;D