Make workers profile aware or profiles worker aware

[adityasaky]

I'd like to be able to use some workers exclusively to rebuild selected profiles, for example, Arch Linux's community repository which has significantly more packages than core and extra. This would also allow for more granular control over which underlying host builds which packages. For example, heavier packages like linux*, gcc etc can be in a separate profile, that is always assigned to a host with the necessary resources.

[AKhozya]

Adding another use case for this feature.

I'm running rebuilderd on a homelab with 2 worker nodes contributing to Arch Linux reproducible builds. The challenge is resource allocation:

The problem:

• Most packages build fine with 6-9GB RAM
• A few monsters (electron, chromium, llvm, rust packages with LTO) need 12-16GB+
• Currently all workers get the same resources, so either:
  • Set high limits (12GB+) → wastes resources on small packages, fewer concurrent workers
  • Set lower limits (9GB) → ~3% of builds OOM and fail

What would help:

• Tag workers with capabilities (e.g., memory=high, memory=low)
• Tag packages or profiles with requirements
• Daemon routes heavy packages to capable workers

Example config (wishlist):

# Worker config
[worker]
name = "worker-small"
tags = ["memory=4G", "cpu=2"]

[worker]
name = "worker-large"  
tags = ["memory=16G", "cpu=8"]

# Package/profile routing
[routing]
packages = ["electron*", "chromium", "llvm*", "rust", "firefox"]
require_tags = ["memory=16G"]

This would allow efficient resource utilization while ensuring heavy builds get routed to capable workers.

Currently working around this by accepting ~3% OOM failure rate and letting packages retry, but proper routing would be much better.

[kpcyrd]

Relevant pull request is #198