This allows spawning a subprocess from a process using lots of memory without issue.
Otherwise the kernel will make sure we have in memeory+swap enough to accomodate 2*size of the parent process.

Made the non-implemented functions of Scheduler throw exceptions instead doing and returning nothing.

Fixed minor issues preventing the queueing of requests in SchedulerTest, DISABLED_archive_to_new_file

Added preliminary implementation of session's HandlerInterface which will communicate with the former. This class will
replace both TaskWatchDog and TapeServerReporter.
A new protocol buffer class (WatchdogMessage) was created to allow communication between the two.
Added the creation of the drive handler in the tape daemon.
Various minor fixes.

Fixed wrong shutdown status reporting in signal handler (it is always ready to shutdown).

Fixed process manager going to polling after timeout handler requested shutdown/kill/fork instead of handling the request.
Gathered the handlers for unit tests in a different file.

The CMakeLists.txt file of the catalogue directory instructs
cmake to generate SqliteCatalogueSchema.cpp, by combining
the contents of the SqliteCatalogueSchema.before_SQL.cpp with the
contents of catalogue_schema.sql.  This means the
SqliteCatalogueSchema.before_SQL.cpp file is not compilable.

The purpose of the SqliteCatalogueSchema structure class is to
isolate the "non-compilable" issues into a small and encapsulated
compilation unit, namely SqliteCatalogueSchema.o, so that the
remaining business logic can be implemented in the non-generated
and compilable SqliteCatalogue.cpp file.  This means that IDEs can
work as normal with SqliteCatalogue.cpp, whereas they will struggle
with SqliteCatalogueSchema.before_SQL.cpp which is therefore
intentionally small.

The SqliteCatalogue.cpp file is now generated by cmake from combining
the SqliteCatalogue.cpp_before_SQL_schema and catalogue_schema.sql
files.