GitLab

We don't want to pay for a double storage (using the `upstream`
namespace) if we don't use it.

We have to use bits of metadata and introduce an explicit
supervision state: `unsupervised`, `supervised` and `handcrafted`.

`handcrafted` comes from manual=True first insertion. Current
semantics should be reviewed.

Also following tshistory 0.8

auto -> upstream
synth -> edited

Dominant use case is to receive things from excel, which
means the manual edition is actually embedded within
mostly "automatic" data. Very different from a carefully
crafted insert ...

So we'll keep the habit of computing the manual insertions
by diffing automatic against the base (synthetic).

tests: make sure we work around https://bitbucket.org/zzzeek/sqlalchemy/issues/4289/automatic-index-name-hash-collision


Fix is in tshistory.

namespaces: make sure the `automatic` and `manual` namespaces are properly related to the base namespace

We build a 'manual' namespace.

This may be useful if we want to use older versions of the original
serie but *still* keep applying the editions over it.

We now don't need the very complicated logic/code we had.

Resolves #1

Unfortunate effect: positional calls have to be fixed.