Move btrfstune functionality to `btrfs tune` subcommand group, and the new interface.

[adam900710]

Currently I have already sent out a patchset to start moving functionality to btrfs tune group.

The eventual planned interface would looks like this:

• btrfs tune set list-all
  To list all supported features to set.

• btrfs tune set block-group-tree <device>

• btrfs tune set no-holes <device>

• btrfs tune set seed <device>
  The feature name follows the one in mkfs.btrfs, the only exception is for seed device.
  All binary features (just simple on and off) would go set/clear subgroup.

For clear it's pretty much the same:

• btrfs tune clear list-all
• btrfs tune clear seed <device>
• btrfs tune clear block-group-tree <device>
  For features like no-holes/free-space-tree they can not be disabled, as they are already default features.

For uuid related (both fsid and metadata_uuid), it would go something like this:

• btrfs tune fsid random <device>
• btrfs tune metadata_uuid random <device>
• btrfs tune fsid 01020304-0506-0708-090a-0b0c0d0e0f10 <device>
  We use random as a special keyword to generate a random uuid/metadata_uuid.

Finally for csum change, it would go its own subgroup command, but mostly follows the same for fsid:

• btrfs tune csum crc32c <device>
• btrfs tune csum sha256 <device>

For resuming the interrupted csum change, just use the same command line as the initial one.
btrfs tune csum would do extra checks just as btrfstune used to do.