Refactor how pads are laid out

[emmakbat]

The current pad ports are always defined by the ports on a PHIDL geometry compass_multi. This is convenient in many ways, but rather inflexible. I'm having two specific problems from this:

1. All pads across one cell will be evenly spaced. This can make it hard to visually distinguish between devices, since e.g. if one device = 4 closely spaced pads, we will just have a uniform row of many equally spaced pads. It would be nice to allow a closer pitch on pads within a device, and more separation for pads belonging to different devices
2. It's hard to set the pitch (see also Issue nanowire cell: align pads and devices #61 to some extent). Compass_multi takes in a size for the compass and a dictionary of how many ports in the north, south, etc. It'll space the pads evenly based on these two parameters.

I'm not sure what the best refactor would be; compass_multi is used pretty elegantly in the code, and in some ways its inflexibility produces nice -- or at least very consistent and understandable -- results. However, it's really not optimal when designing for a specific probe pitch. Maybe we should keep the compass_multi behavior for devices intended for wirebonding and have alternate behaviors for probe station tips.

[emmakbat]

Better covered by issue #111