port ftp slave to new error reporting system

Summary:
This system is a bit inspired by an earlier patch I threw together in
a half hour to prevent the actual logic code from calling state-changing
functions by splitting the interface class (implements SlaveBase) from
the actual FTP client implementation (used by the interface class).
The advantage here is that the future internal code cannot call error
or finished directly (and cause state violations in SlaveBase - e.g.
double-error call) because there is no trivial way to access them.

Instead the internals return a new Result class if they need to return
anything. This is largely inspired by how Error handling works in Go.
The caller of an internal function needs to check the result or explicitly
disregard it. The caller may then decide to ignore an internal error result
(e.g. chmod after upload failed), create its own higher level result, or
forward it unchanged. On the interface level we then convert the result
to appropriate state changes of the slave.
This is assisted by Q_REQUIRED_RESULT so capable compilers will warn if
a Result is not explicitly ignored as that'd be an indication of missing
error handling in new code, again future proofing the slave.

All things put together this turned out really well I think.

This also adds a primitive test because it broke a million times while
porting. Starts a simple ruby ftpd (with some monkey patching applied).

Reviewers: dfaure

Reviewed By: dfaure

Subscribers: anthonyfieroni, dfaure, kde-frameworks-devel

Tags: Frameworks

Differential Revision: https://phabricator.kde.org/D23579