Refactor shared code between latency and latencybg tests

[arlake228]

Currently the latency and latencybg tests have a lot of copy and pasted code between them which make it challenging when you need to do updates since you have to do it in two place. Its made worse by the fact that its clearly copy and pasted but file structures and such don't even match, so its "just close enough to be annoying". For example, the result formatting is almost exactly the same even though each test has its own independent copy. Recently we fixed how bucket width is handled in latency and that did not make its way to latencybg since it was separate even though the code is similar.

We should give them some proper shared libraries between them for code like this. They are legitimately different tests, but for the common bits we should only need to update code in one place.

[mfeit-internet2]

Semi-related: #994

[mfeit-internet2]

The plan for this should be:

• [ ] Adjust the internal architecture so the winning tool plugin selects the scheduling class instead of having it hard-wired in the test enumeration. This won't matter because no scheduling happens until the tool has been selected.

• [ ] Merge latencybg into latency, adding parameters that will allow tools to accept/reject tasks depending on the parameters. (E.g., owping only accepts tests without parameters that make them long-running and powstream accepts those with them.)