Add storage API

[gibmat]

Adds a preliminary storage API:

• /1.0/system/storage:
  • GET returns details about each drive and the status of each zpool
  • PUT creates/updates a new zpool
  • DELETE destroys a zpool and runs blkdiscard on its former members
• /1.0/system/security:
  • GET also returns base64 encoded zpool encryption keys

Closes #145

[stgraber]

It'll be more work, but I think we should try to abstract things a bit more.

Basically having a Drives []SystemStorageDrive with each drive then including at least:

• ID
• Manufacturer
• Model
• Serial number
• Bus
• Removable
• WWN
• SMART status
• Pool

We'll also want to do a bit of filtering.
For example, if we find a multipath device, we should only show the ID for the multipath device itself and exclude all of the individual drives that are part of it.

[gibmat]

I think I've got a pretty good solution for this, although it will be interesting to see if any fields are missing/incorrect across wider hardware.

lsblk has a merge option (-M), which supposedly should handle cases like multipath devices and only show the actual device, but we'll need to test that.

[stgraber]

I think defining and updating pools should be done through the Config struct same we do elsewhere.

So having something like:

Config {
    Pools [
        Pool {
            Name: "local-nvme",
            Type: "zfs-raidz1",
            Devices: ["dev1", "dev2"],
        }
        Pool {
            Name: "local-nvme",
            Type: "zfs-raidz2",
            Devices: ["dev1", "dev2", "dev3", "dev4", "dev5"],
            Cache: ["dev6", "dev7"],
            Log: ["dev8", "dev9"],
        }
    ]
}

We'd define those types for now:

• zfs-raid0 (stripe across all devices)
• zfs-raid1 (mirror across all devices)
• zfs-raid10 (stripe across pairs of devices)
• zfs-raidz1
• zfs-raidz2
• zfs-raidz3

[stgraber]

Replacement operations would then literally be a PUT to the config, replacing one drive ID with another. Same would go for device addition or removal (when supported).

[stgraber]

Looks like there's a small test migration related to the state

[stgraber]

Static analysis isn't happy ;)