Scale lab queue limits based on number of devices

Since the introduction of support for limiting the number of jobs in https://github.com/kernelci/kernelci-core/commit/aeb2684203484f69447aacf903f5b7f09b19c138 we queue on labs we've used an absolute limit on the number of jobs that's specified as a simple number of jobs. At the minute that's 50. This limit is applied no matter how many devices that a lab has available, so for example with my lab where I have 5 BeagleBone Blacks available with average job runtimes of approximately 5 minutes we'd clear 50 jobs in about an hour which is reasonably responsive. In contrast if I only had one board the same jobs would take over 4 hours which is much less usable.

A simple improvement would be to scale the limit based on the number of devices of a given type that a lab has, picking a smaller limit per board but allowing us to make use of larger lab capacities.

Start with commit aeb2684203484f69447aacf903f5b7f09b19c138 and trace where the current absolute lab queue limit is applied. Review how device types and available lab capacity are represented, then define the scaled limit and verify that labs with different numbers of devices receive appropriate queue limits.

Written by the indexing model from the issue text.