In ReplicationReport, remove attributes that correspond to statistics of
individual that can be recomputed by reading the genome. A major problem
was also when recording Trees, if statistics was not computed, the
values are completely wrong. Finally, checking if execution is
repeatable, the contant of tree files might change according to a spooky
and hardly predictable state (AEVOL_NO_STATS)

Also some reformat in includes

+ Turn List_Metadata as a standalone class
+ Remove all others obsolete inheritors that did not even compile
+ Propagate type usage in multiple files
+ Un-inline some big methode of metadata
+ move definition of constant from ExpManager_7.h to macros.h

It is now possible to choose the selection mode (global or local) for
the run. Local is using a default scope size.

+ Add global selection feature
+ Create an abstract class Selector that allow to implement different
selection routine
+ make LocalSelector implement Selector
+ Create GlobalSelector that implement Selector
+ Add an attribute Selector in ExpManager_7 that will call the right
selection routine. This attribute is set during construction

In module_run.cpp, an old and not used define wrongly set NB_BASE at 2.
In practice, it does not change anything, but it was misleading

+ Put selection routine out of ExpManager_7 inside a new file. Keep the
code of the old one for prosperity
+ LocalSelector is then a functor parametrized by selection scope and
grid dimension
+ The new routine only keeps what is relevant for local selection,
without MPI or regulation features

+ Put translation of protein routines in separate files
+ Create functions for 2 and 4 bases
+ Use translation function as function "pointer/object"
+ For 4 bases, using a constexpr functor as there are parametrized data
for the translation

Take out some 4 bases feature from `#define` guards.
The goal is to eventually put them all in separate files

`bu::write_backup` now use const parameter as it should only read data
and never change them.

Stop using the variable `is_indiv` (while writing) in Stats_7, that
record a state that change the behavior of the object in a hardly
predictable manner

In Individual_7, this new methode allow to evaluate completely an
individual that has just mutate. It means that a part of its metadata is
kept to accelerate computations

In Dna_7, when v6 mutations (Mutation) are used, the mutation counters
are correctly updated as when using v7 mutation (MutationEvent)

Metadata are properly reset in evaluate_from_scratch()

+ Remove a static function from ExpManager_7 to put it in Stat_7

+ Use type aliases to ease usage
+ remove usage of ExpManager to remove constraint
+ modify run in consequence

+ Create a genome generator, that read json and spit out a json with a
genome using an environment
+ Create a runner that begins with a clonal population. It can produce a
tree readable by lineage

+ Add functions that save and load state of experiments (genome and
prngs state)
+ TODO: choose better names for backup and put them in a better place

+ Remove the responsibility of the v7 constructor for the creation of
its data state, i.e. population and prngs
+ Creating the population and the prngs has to be done before running
the evolution
+ The first evaluation of the individuals has to be done outside
+ The goal is to create specific services of population and prngs

+ Create a way of creating an experiment completely without v6
+ v6 pointer is set to NULL.
+ Backup using v6 is done only if v6 is not NULL
+ Only works with clonal population

No change in the affect but separate clearly the gathering of all
parameters and the construction of the object

+ evaluate_from_scratch() has a very misleading name. The state of
metadata is not completely cleared before evaluation

+ Backup is still not done

+ Create only one instance of the same indiv for clonal population

+ Use vector of shared pointers to store individuals
+ Remove custom memory management for individual, smart pointers are
there to do it
+ Set first population with only previous_individuals. It can be renamed
current_population, whereas current_individuals refers more for the
next_population
+ WARNING! best indiv is not updated automatically. Some printing have
then no meaning. Moreover, this indiv will never be deleted.

+ remove exp_manager from tree
+ creation of tree does not use ExpManager
+ It still has a dependence on ExpManager through Individual (v6)
+ Add indiv_id to EndReplicationEvent to use it in
Tree::update_end_replication()
+ Propagate changes in signature through many files
+ v7 now use its own Tree without v6

+ Turn FuzzyFactory into a Factory-ish only. No more allocator features
+ FuzzyFactory create concrete Fuzzy pointed by an AbstractFuzzy
+ Memory is to be manage by the user

+ Memory of Dna_7 object is managed with `new` and `delete`
+ Change signature of methods
+ Propagate changes in multiple files