[blueprint-planner] Don't fail on sleds with no zpools

[jgallagher]

Fixes #7800. That issue notes that this is a minor inconvenience when adding a sled, but (a) it's easy to fix and (b) we hit it in colo in a different context (expunging all the disks on a sled). That combo made it seem worth it to go ahead and fix it.

[jgallagher]

I'll take this branch for an a4x2 run before merging to make sure having this empty config in the blueprint doesn't cause execution problems, but I think it should be fine.

[jgallagher]

Seems fine on a4x2. Added a sled, generated a blueprint that has an empty config:

root@oxz_switch:~# omdb nexus blueprints show current
...
  sled: 9a64bc86-2b6e-4299-bf61-13702a58bce5 (active, config generation 1)

    physical disks:
    -------------------------------------
    vendor   model   serial   disposition
    -------------------------------------


    datasets:
    ---------------------------------------------------------------------------
    dataset name   dataset id   disposition   quota   reservation   compression
    ---------------------------------------------------------------------------


    omicron zones:
    --------------------------------------------------------------
    zone type   zone id   image source   disposition   underlay IP
    --------------------------------------------------------------
...
 METADATA:
    created by:::::::::::::   b4ab66ba-c08b-461a-9dbd-5c90e0e180f3
    created at:::::::::::::   2025-06-17T18:32:25.600Z
    comment::::::::::::::::   skipping sled 9a64bc86-2b6e-4299-bf61-13702a58bce5: no zpools in service

Executing it shows no errors or warnings on the Nexus side, and resulted in the empty config being ledgered on the sled:

root@g2:~# cat /pool/int/1d8bafd3-bb55-4461-8fcf-8b9c91cd59b2/config/omicron-sled-config.json
{"generation":1,"disks":{},"datasets":{},"zones":{},"remove_mupdate_override":null}

Generating another blueprint later was also fine - we added disks and datasets and an NTP zone, and that config was ledgered correctly.