Make the set of weekly runs configurable from `bench_runner.toml`

[mdboom]

Currently, the set of weekly runs is defined in _weekly.src.yml using GitHub actions. It currently does:

• default builds (all machines)
• jit builds (all machines)
• nogil builds (all machines)
• tailcall builds (all machines)
• pystats builds (linux "cloud" machine)

With #396, (and probably other things), there may be certain combinations of flags and runners that don't work, or just cases where the matrix is so large we don't want to run all of it.

I would propose moving the configuration of the weekly runs to bench_runner.toml where the weekly runs are defined as multiple products of flags and runners. For example:

[weekly]
default = {
    flags: []
    runners: [linux, darwin, windows]  # runner nicknames
}
nogil = {
    flags: [NOGIL],
    runners: [linux, darwin, windows]
}
tailcall = {
    flags: [TAILCALL],
    runners: [linux-clang19, darwin, windows-clangcl]
}

This information would be used by the install.py script to produce a weekly.yml for GitHub Actions to run.

@Yhg1s, @Fidget-Spinner: Your thoughts?

[Yhg1s]

I have so far ignored automatic runs in my own repo since I just have too many configurations for GHA to run them all (GHA will only wait for 24h for a runner to become available), but yes, I think it would be great to have it configurable. I'm still whittling down the list of configurations that actually work in my setup, but I will definitely end up with too many to run automatically. I would probably only configure one or two configurations to run weekly, and have some other mechanism take care of running more configurations at the same commit during the rest of the week.. so as long as it can be configured or turned off, I'm happy.