feat: add autoscale settings to datumctl compute

[savme]

This PR adds CPU/memory-based autoscaling to deploy and scale in datumctl, and surfaces it in workloads describe.

Deploy with autoscaling from the start:

datumctl compute deploy api --image=ghcr.io/acme/api:1.4.2 --location=us-east-1 --min=1 --max=3 --cpu-percent=70

Turn autoscaling on or adjust it on an existing workload, without redeploying:

datumctl compute scale api --max=5 --cpu-percent=70
datumctl compute scale api --max=8 # just raise the ceiling, autoscaling stays as configured
datumctl compute scale api --max=0 # turn autoscaling back off

See what's configured:

datumctl compute workloads describe api

Placements
  default    us-east-1, eu-west-1              scale: 2..10 (cpu@70%)

Notes:

• Flags only change what was passed - running scale api --max=5 on an autoscaled workload doesn't reset its CPU target
• If a workload was declared with multiple named placements via a manifest, deploy warns before collapsing them into one (flags produce only a single placement)

Test plan

• Unit tests pass
• datumctl compute scale and datumctl compute deploy tested