This beefy commit removes the old (unnamed) Servant API.

This big commit ties the knot with the named _routes_ by providing
concrete server (i.e. handlers) implementations.

It forces programmers to think about errors we are logging and reporting
to the frontend, because they need to contain no sensitive data.

This contains only data, removes some Database dependency in imports.

Also, some test refactoring and add servant-client to tests.

Get rid of that TVar completely.

Added more tests for more subtle cases.

Also, some fixes to NodeStory (it inserts ngrams when saving node
story, when they don't exist).

Basically, Gargantext.Prelude exports all of Protolude now.

FlowCmdM -> DBCmd or HasNodeStory is usually enough

The issue was that when search was performed, the initial table didn't
have an archive ready and so first commit resulted in some errors on
the frontend. This fix performs the commit after flow is finished.

This commit fixes a bug (or feature? unclear at this stage) of the
`rootOf` function inside `searchTableNgrams`, which could, in some
occasion, return a root which has a different `ListType`, which
in turn throws off pagination.

This commit adds some basic tests for the Ngrams pagination,
capitalizing on the refactoring of the `searchTableNgrams` function,
which is now pure and can be tested fairly easily.

This commit refactors the Ngrams pagination code to use
the `NgramsSearchQuery`, and it introduces a new function called
`searchTableNgrams`, which is now completely pure.

This is safer than type Limit = Int

This big commit improves the jobs API in a way that now we can
completely abstract away over a JobLog.

Related to gargantext/purescript-gargantext#441

We had a conflict here: archive history is used to resolve new
patches. However, they were saved with delay and after that were
cleared. We should save archive immediately and immediately clear it
in mvar so it's ready for new patches.