Allow to use an existing snapshot at upgrade time

For the peace of mind of user and dev, 66 handles now
on its own the migration of the different components
of the 66 ecosystem and user or distributor have nothing to do at upgrade time.
The migration process do not affect running services.
Its also make a snapshot before making the migration. So, user
can retrieve easily the previous working state of the system in case
of crash of the migration process.

Better errno handling

the version of the system and the version of services

create, restore, remove and list are available.
These subcommand do not affect the state of a running services.

Add islog internal key to quickly is the resolve file concern the logger itself.
Add the rversion internal field. This field will be used to know the previous
version of the resolve file and possibly make an
automatic migration of the db to the new format.

Typo fix

Follow skarnet changes using stralloc and genalloc.

In addition to the string "@I" as identifier to replace by the name of instance service,
the following identifier are now recognize :
@I, @U, @u, @G, @g, @H, @S, @R. see doc/66-identifier.md for furthers information.

See commit 3f345ecb
Rename tree_resolve_copy and tree_resolve_master_copy to
tree_resolve_sanitize and tree_resolve_master_sanitize respectively.

At write time:
  - Add the whole sa to the new cdb field named sa instead of storing the key one by one.
  - Store uniquely the uint of the key instead of appending a string.

At read time:
  - Do not build again the sa looking for the cdb value but read the whole sa instead.
  - Do not use extra HEAP memory allocation.
At modification time:
  - Sanitize the resolve by writting the sa from scratch.
  - Use /tmp directory as temporary cdb file.
Replace resolve_find_cdb by the resolve_get_key and resolve_get_sa function.
Replace service_resolve_copy by service_resolve_sanitize.