Combine inv_sled_omicron_zones with inv_sled_agent

[sunshowers]

Followup from #6739.

Previously, the collection process would gather Omicron zones and inventory from sled-agent in separate HTTP requests. Since they can fail independently, information about them was recorded separately in two tables.

Omicron zone information is recorded in inv_sled_omicron_zones:

omicron/schema/crdb/dbinit.sql

Lines 3331 to 3338 in 81b327d

	CREATE TABLE IF NOT EXISTS omicron.public.inv_sled_omicron_zones (
	-- where this observation came from
	-- (foreign key into `inv_collection` table)
	inv_collection_id UUID NOT NULL,
	-- when this observation was made
	time_collected TIMESTAMPTZ NOT NULL,
	-- URL of the sled agent that reported this data
	source TEXT NOT NULL,

And general inventory is recorded in inv_sled_agent:

omicron/schema/crdb/dbinit.sql

Lines 3199 to 3206 in 81b327d

	CREATE TABLE IF NOT EXISTS omicron.public.inv_sled_agent (
	-- where this observation came from
	-- (foreign key into `inv_collection` table)
	inv_collection_id UUID NOT NULL,
	-- when this observation was made
	time_collected TIMESTAMPTZ NOT NULL,
	-- URL of the sled agent that reported this data
	source TEXT NOT NULL,

---

In #6739, we merged the two into a single HTTP request, but for convenience reasons we still record them in separate tables.

This task tracks cleaning this up so that the two are in the same table.

---

The only important column to preserve from inv_sled_omicron_zones is generation. Ideally this would be added to inv_sled_agent as a non-null column.

The easiest way to do this would be to drop all old collections then add the generation column. But doing that across a version upgrade can leave us in a fragile state with no collections available. So it's probably better to do an inner join by collection ID, so most collections are preserved.